Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
PRODUCT RECOMMENDATION SYSTEM AND METHOD
Document Type and Number:
WIPO Patent Application WO/2020/053307
Kind Code:
A1
Abstract:
A wearable device comprising: a memory configured to store product codes for consumable, topically applied and/or body-worn products and data indicating respective product recommendations or from which product recommendations can be derived; a product code reader for reading product codes from products; one or more inertial sensors for obtaining motion data for a wearer of the device; a visual indicator for providing a visual indication of a product recommendation, using data stored in the memory, in response to a read product code. The wearable device further comprises a processor configured to process the motion data to identify periods when the wearer is in a sitting position or other sedentary state, analyse the occurrence and durations of the periods, and modulate the recommendations accordingly for at least a subset of the product codes, whereby product recommendations change depending upon the identified periods.

Inventors:
KARVELA MARIA (GB)
TOUMAZOU CHRISTOFER (GB)
Application Number:
PCT/EP2019/074277
Publication Date:
March 19, 2020
Filing Date:
September 11, 2019
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
DNANUDGE (GB)
International Classes:
G16H20/60; A61B5/11; G16H20/70; A61B5/00
Domestic Patent References:
WO2017055867A12017-04-06
Foreign References:
US20030208110A12003-11-06
US20160071423A12016-03-10
US20030226695A12003-12-11
EP2416269A22012-02-08
US20170098268A12017-04-06
Other References:
EKELAND ET AL., THE LANCET, vol. 388, 2016, pages 1303 - 1310
Attorney, Agent or Firm:
WHITE, Duncan (GB)
Download PDF:
Claims:
CLAIMS:

1 . A wearable device comprising:

a memory configured to store product codes for consumable, topically applied and/or body-worn products and data indicating respective product recommendations or from which product recommendations can be derived;

a product code reader for reading product codes from products;

one or more inertial sensors for obtaining motion data for a wearer of the device;

a visual indicator for providing a visual indication of a product recommendation, using data stored in the memory, in response to a read product code; and

one or more processors configured to process the motion data to identify periods when the wearer is in a sitting position or other sedentary state, analyse the occurrence and durations of said periods, and modulate said recommendations accordingly for at least a subset of said product codes, whereby product recommendations change depending upon the identified periods.

2. A wearable device according to claim 1 , wherein the one or more processors are configured to modulate recommendations according to the wearer’s behaviour determined by the processor for a pre-determined period, such as the same day.

3. A wearable device according to claim 1 or 2, wherein the one or more processors are configured to identify and maintain a count of periods when the wearer is in a sitting position or other sedentary state for a duration longer than a threshold duration, such as 10 minutes, 30 minutes or 60 minutes.

4. A wearable device according to any one of the preceding claims, wherein the one or more processors are configured to ignore interruptions in which the user is not in a sitting position or other sedentary state for less than a pre-determined time when identifying said periods.

5. A wearable device according to any one of the preceding claims, wherein said memory stores nutritional data for each product and said processor is configured to select said subset of product codes based on said nutritional data.

6. A wearable device according to any one of the preceding claims, wherein each product recommendation has a first, recommended state and a second, not recommended state and said modulation changes a product recommendation between the first and the second state.

7. A wearable device according to any one of the claims 1 to 5, wherein said product recommendation has a first, recommended state, a second “maybe recommended” state and a third, not recommended state and said modulation changes the product recommendation between the first and the second state.

8. A wearable device according to any one of the preceding claims, wherein the wearable device is a wrist-worn wearable device.

9. A wearable device according to any one of the preceding claims, wherein said one or more inertial sensors comprises an accelerometer and/or a gyroscope.

10. A wearable device according to any one of the preceding claims, wherein said product code reader is for reading a one-dimensional or two-dimensional barcode.

1 1 . A wearable device according to any one of the preceding claims, wherein said visual indicator provides said indications of the product recommendations by means of different coloured illuminations, for example red and green, or red, green and amber.

12. A wearable device according to any one of the preceding claims, wherein the one or more processors are configured to, based on the behaviour of the user, select from at least a subset of the product codes, product codes for which the respective product recommendation is to be modulated; and wherein the or another visual indicator is controllable by the one or more processors to provide another visual indication, the other visual indication being dependent on the number of selected products codes.

13. A method of providing recommendations in respect of consumable products to a wearer of a wearable device in order to improve the wearer’s health, the method comprising: storing, in a memory of the wearable device, product codes for products and data indicating respective product recommendations or from which product recommendations can be derived;

obtaining, from one or more inertial sensors of the wearable device, motion data for the wearer;

processing the motion data using a processor of the wearable device to identify periods when the wearer is in a sitting position or other sedentary state;

analysing the occurrence and durations of said periods using said processor;

reading a product code from a product using a product code reader of the wearable device;

in response to a read product code, using data stored in said memory to obtain a product recommendation;

modulating said product recommendation depending upon the wearer’s sitting behaviour, or other sedentary behaviour, determined by the processor; and

providing a visual indication of the modulated product recommendation using a visual indicator of the wearable device.

14. A method according to claim 13, further comprising selecting a subset of product codes for which the associated product recommendations are modulated.

15. A method according to claim 14, wherein selecting a subset of product codes for which product recommendations are modulated comprises selecting products codes based on nutritional information for the associated products.

16. A method according to claim 15, wherein selecting products codes based on nutritional information for the associated products comprises ranking the products based on the respective nutritional information for each product and selecting the subset by selecting products having a ranking higher than a pre-determined ranking.

17. A method according to claim 15 or 16, wherein the nutritional information comprises one or more of: calorific content, sugar content, carbohydrate content, saturated and/ or unsaturated fat content, and salt content,

18. A method according to any one of claims 13 to 17, wherein said data is derived from personalised data derived from personalised biological information obtained from an analysis of a biological sample provided by the wearer of the device.

19. A method of providing product recommendations in respect of consumable products in order to benefit the health of a user, the method comprising:

identifying one or more health traits of the user;

for each of a multiplicity of consumable, topically applied and/or body-worn products, calculating a score indicative of how the product affects each of said health traits, each score being based at least partly on nutritional information for the product;

assigning a product recommendation to each product based on said scores;

monitoring the behaviour of the user by identifying periods when the user is in a sitting position or other sedentary state;

based on said behaviour of the user, modulating the product recommendation for at least of a subset of said products; and

providing the modulated product recommendation to the user via a visual indicator.

20. A method according to claim 19, wherein the product recommendation for each product is stored in a database installed in a memory of a wearable device, said memory further storing data indicative of whether the product belongs to said subset.

21 . A method according to claim 20, wherein the wearable device comprises one or more inertial sensors for identifying periods when the user in a sitting position or other sedentary position.

22. A monitoring system for alerting a user to negative effects caused by a user’s environment and/or lifestyle, the monitoring system comprising:

a memory storing product codes for consumable, topically applied and/or body-worn products, and data indicating respective product recommendations or from which respective product recommendations can be derived;

one or more processors configured to:

monitor the behaviour of the user using data indicative of one or more physiological and/or biochemical functions of the user, or indicative of a user environment, the data being obtained using one or more user-worn sensors; and based on the behaviour of the user, select from at least a subset of the product codes, product codes for which the respective product recommendation is to be modulated; and

a visual indicator controllable by the one or more processors to provide a visual indication dependent on the number of selected products codes.

23. A monitoring system for alerting a user to the negative effects of sedentary behaviour, comprising:

a memory storing product codes for consumable, topically applied, and/or body-worn products, and data indicating respective product recommendations or from which respective product recommendations can be derived;

one or more processors configured to:

monitor the behaviour of the user by using motion data obtained from one or more inertial sensors worn by the user to identify periods when the user is in a sitting position or other sedentary state; and

based on the behaviour of the user, select from at least a subset of the product codes, product codes for which the respective product recommendation is to be modulated; and

a visual indicator controllable by the one or more processors to provide a visual indication dependent on the number of selected products codes.

24. A monitoring system according to claim 23, wherein the one or more processors are configured to determine a count of the number of periods when the user is in a sedentary state for longer than a threshold amount of time and to increase the number of selected product codes when said count increases.

25. A monitoring system according to claim 24, wherein the one or more processors are configured to increase the number of selected product codes by successively greater amounts as said count increases.

26. A monitoring system according to claim 24 or 25, wherein the one or more processors are configured to use the received motion data to determine an amount of physical activity performed by the user and to reduce the count of the number of periods based on the amount of physical activity performed by the user.

27. A monitoring system according to claim 26, wherein the one or more processors are configured to determine, from the motion data, the number of steps taken by the user.

28. A monitoring system according to any one of claims 24 to 27, wherein the one or more processors are configured to apply the modulation to the product recommendations only if the count exceeds a pre-defined limit.

29. A monitoring system according to claim 28 and comprising a user interface for receiving user input and wherein the processor is configured to adjust the limit based on the user input.

30. A monitoring system according to any one of claims 23 to 29, wherein each product recommendation has a first, recommended state and a second, not recommended state and said modulation changes a product recommendation between the first and the second state.

31 . A monitoring system according to any one of claims 23 to 29, wherein each product recommendation has a first, recommended state, a second “maybe recommended” state and a third, not recommended state and said modulation changes the product recommendation between the first and the second state.

32. A monitoring system according to any one of claims 23 to 31 , wherein said visual indicator is configured to display a graphical element having a length or area representative of the number or proportion of selected products codes.

33. A monitoring system according to any one of claims 23 to 32, wherein the system is configured to receive a product code, retrieve or derive a product recommendation for the consumable product associated with the product code, and to modulate the product recommendation if and only if the product code is one of the selected product codes.

34. A monitoring system according to claim 33, and comprising a product code reader for reading a product code from a product.

35. A monitoring system according to any one of claims 23 to 34, wherein the processor is configured to select the product codes for which the respective product recommendation is to be modulated at least in part based on nutritional information for the associated products.

36. A monitoring system according to any one of claims 23 to 35, wherein the one or more processors are configured to rank the products based on the respective nutritional information for each product and select the product codes by selecting products having a ranking higher than a pre-determined ranking.

37. A monitoring system according to claim 36, wherein the nutritional information comprises one or more of: calorific content, sugar content, carbohydrate content, saturated and/ or unsaturated fat content, and salt content.

38. A monitoring system according to any one of claims 23 to 37, wherein said data is derived from personalised biological information obtained from an analysis of a biological sample provided by the user.

39. A monitoring system according to any one of claims 23 to 38, wherein the visual indicator is a display of a personal computing device, such as a smartphone or smartwatch.

40. A wearable monitoring device for alerting a user to the negative effects of sedentary behaviour, comprising:

a memory storing product codes for consumable, topically applied and/or body-worn products, and data indicating respective product recommendations or from which respective product recommendations can be derived;

one or more inertial sensors for obtaining motion data for a wearer of the device;

one or more processors configured to:

monitor the behaviour of the user by using the motion data to identify periods when the user is in a sitting position or other sedentary state; and

based on the behaviour of the user, select from at least a subset of the product codes, product codes for which the respective product recommendation is to be modulated; and a visual indicator controllable by the one or more processors to provide a visual indication dependent on the number of selected products codes.

41 . A method for alerting a user to the consequences of sedentary behaviour, the method being implemented by one or more computer devices and comprising:

storing in a memory of at least one of the computer devices, product codes for consumable products and data indicating respective product recommendations or from which respective product recommendations can be derived;

obtaining motion data from one or more inertial sensors worn by the user;

monitoring the behaviour of the user by using the motion data to identify periods when the user is in a sitting position or other sedentary state;

based on the behaviour of the user, selecting from at least a subset of the product codes, product codes for which the respective product recommendation is to be modulated;

controlling a visual indicator to provide a visual indication dependent on the number of selected product codes.

42. A method according to claim 41 , wherein the selected product codes are selected from a subset of the product codes for which the respective product recommendations are eligible for modulation and further comprising selecting said subset based on nutritional information for the associated products.

43. A method according to claim 42, further comprising ranking the products based on the respective nutritional information for each product, wherein selecting the product codes comprises selecting products having a ranking higher than a pre-determined ranking.

Description:
PRODUCT RECOMMENDATION SYSTEM AND METHOD

Technical Field

The present disclosure relates to wearable devices and methods for providing recommendations of products. It also relates to monitoring systems, wearable devices, and methods for alerting a user to the negative effects of caused by a user’s environment and/or lifestyle, such as sedentary behaviour. In particular, but not exclusively, the present invention relates to providing recommendations based on product content and consumers’ personal biological information.

Background

Semiconductor nanotechnology and optical technologies have made significant contributions to people’s lifestyle, especially by facilitating hardware miniaturisation. Its application to the sequencing and genotyping industry has enabled so-called“lab-on- chip” systems. Depending on the biological questions/genes of interest, primer(s)/probe(s) - more generally referred to as “biomarkers” - are designed accordingly. A biomarker is an oligonucleotide such as a DNA molecule and may target certain gene(s)/variation(s). A biomarker may alternatively, for example, be an antibody or an antigen. By applying/choosing different types of biomarkers on such systems, a customer can test his/her biological sample, DNA, RNA, protein etc, (extracted locally or remotely by a third party from e.g. saliva, blood, urine, tissue, stool, hair etc) for specific traits, possibly as dictated by certain lifestyle concerns or interest.

Such“personal” genetic or biological information enables medical decisions to be made more effectively, for example, by selecting treatments or drug doses which are more likely to work for particular patients. Identifying individual differences at a molecular level also allows lifestyle and dietary advice to be tailored according to the needs of individuals or particular classes of individuals. For example, personal care products such as cosmetics and nutraceuticals may be selected based on how effective these products are for individuals having certain single nucleotide polymorphisms in their DNA. A number of private companies have been created in order to cater for the growing consumer genetics market and every day new genetic traits are being described, generating a continuously expanding catalogue of biomarkers that have the potential to offer insight into the health, wellbeing, and, in the case of genetic variations, phenotype, of a great many people.

Personal activity monitors provide users with a convenient way of recording their physical activity. In particular, many so-called“fitness trackers” provide users with estimates of the distance they have walked or run, or the total energy they have expended and so forth. This kind of data allows users to make more informed decisions regarding their lifestyle. In some cases, a fitness tracker may remind the user to perform that they have not undertaken moderate or vigorous physical activity recently. While such fitness trackers may help individuals improve their health by motivating them to undertake more physical activity, other factors such as the food and drink the user consumes also play an important role in determining the wellbeing of the individual.

WO2017055867 describes a wearable device for providing product recommendations based on a user’s biological information, such as genetic data. The wearable device incorporates a laser scanner or barcode reader, which the wearer of the device uses to identify a product he or she is interested in purchasing or consuming. The device then provides an indication whether or not the product is recommended for the wearer based on his or her biological information. For example, an analysis of a user’s DNA may have revealed that the user metabolises caffeine more slowly than most other people, in which case, the wearable device may recommend that he or she avoids coffee.

The effectiveness of product recommendations based on a user’s biological (genetic) and/or physiological information in providing health benefits to the user can vary depending on the behaviour of the user. There is therefore a need to improve the effectiveness of product recommendations in order to improve the health of users.

Summary

Aspects of the invention are set out in the independent claims. Other aspects and preferred features are set out in the dependent claims. The term“consumable” is used in this document principally to identify products that are consumable orally, e.g. food, beverages, supplements, medicines, etc, although it also encompasses products that are consumed through the skin. The term“topically” is used to mean applied externally to the body, for example, to the skin or hair. Examples of topically applied (or appliable) products include cosmetics, creams, powders or fluids.

Also described herein is a computer-implemented system for providing recommendations to a user in respect of consumable or topically applied (i.e. topically “applicable”) products, at least a part of which system is a body-worn part. The system comprises:

a data store for storing product codes and data derived from personalised biological information obtained from an analysis of a biological sample provided by the user;

a reader for reading or otherwise obtaining product codes from products or product packaging;

a processor for using the read or otherwise obtained product codes, and data stored in said data store including at least said data derived from personalised biological information, to obtain product recommendations for the products;

a user interface for providing indications of the product recommendations to the user; and

one or more sensors for obtaining data indicative of one or more physiological and or biochemical functions of the user, or indicative of a user environment, wherein said processor is configured to modulate one or more of said product recommendations in dependence upon the obtained data whereby a modified indication is provided to the user via said user interface.

At least the reader and the one or more sensors may be provided on said body-worn part, or the entire system may be a body-worn part. The body-worn part may be a wrist-worn part including a wristband. The one or more sensors may comprise an accelerometer.

The processor may be configured to determine, from data provided by the accelerometer, a value indicative of user activity, for example a step count, said modulation of one or more of the product recommendations being dependent upon that activity value. The modulation may be applied to product recommendations taking into account a calorific content of the products. The one or more of the sensors may comprise one or more of a gyroscope, a heart rate monitor, a body fluid or chemical sensor optionally comprising micro-needles.

The personal biological information may be personal genetic information.

The processor may be configured to store in the data store a historical record, for example encompassing a predefined preceding time period, of the data indicative of one or more physiological and or biochemical functions of the user and to use that record to modulate one or more of said product recommendations so that the modulation takes into account a history of the one or more physiological and or biochemical functions of the user.

The data store may be configured to store information relating to product content including, for example, carbohydrate and or sugar product content amounts.

The product recommendation may have a first, recommended state and a second, not recommended state and said modulation changes the product recommendation between the first and the second state. The product recommendation may have a first, “recommended” state, a second, “not recommended” state, and a third “may be recommended” state, and said modulation changes the product recommendation between the first and the third state.

The user interface may be configured to provide the indications of the product recommendations by means of different coloured illuminations, for example red and green, or red, green and amber.

The computer-implemented system may comprise a further user interface for receiving a modulation value from a user, said processor being configured to scale said modulation of the one or more of said product recommendations in dependence upon said modulation value.

The reader may be a barcode scanner. A body worn computer-implemented system for providing recommendations to a user in respect of consumable or topically appliable products is also described. The system comprises:

a data store for storing product codes and data derived from personalised biological information obtained from an analysis of a biological sample provided by the user;

a reader for reading or otherwise obtaining product codes from products or product packaging;

a processor for using the read or otherwise obtained product codes, and data stored in said data store including at least said data derived from personalised biological information, to obtain product recommendations for the products;

a user interface for providing indications of the product recommendations to the user; and

one or more sensors for obtaining data indicative of one or more physiological and or biochemical functions of the user, or indicative of a user environment, wherein said processor is configured to modulate one or more of said product recommendations in dependence upon the obtained data whereby a modified indication is provided to the user via said user interface.

The body worn computer-implemented system may comprise a wristband.

A computer-implemented method for providing recommendations to a user in respect of consumable or topically applied (i.e. topically applicable) products is also described. The method comprises:

storing in a data store, product codes and data derived from personalised biological information obtained from an analysis of a biological sample provided by the user;

reading or otherwise obtaining product codes from products or product packaging;

using the read or otherwise obtained product codes, and data stored in said data store including at least said data derived from personalised biological information, to obtain product recommendations for the products;

providing indications of the product recommendations to the user via a user interface; and obtaining from one or more sensors data indicative of one or more physiological and or biochemical functions of the user, or indicative of a user environment, and modulating one or more of said product recommendations in dependence upon the obtained data whereby a modified indication is provided to the user via said user interface.

Although the one or more sensors referred to above are for measuring one or more physiological functions of the users, alternatively, or as well as, the system or device may comprise one or more sensors for determining the location of the user or one or more environmental factors that the user is exposed to, such as a level of pollution (e.g. NOx or particulates) or ultraviolet (UV) light levels.

Also described herein is a computer-implemented system for providing recommendations to a user in respect of consumable or topically appliable products, at least a part of which system is a body-worn part. The system comprises:

a product code reader;

one or more sensors for obtaining data indicative of one or more physiological and or biochemical functions of the user, or indicative of a user environment;

a processor configured to determine product recommendations for products identified using the product code reader, based upon a user’s personal biological information and data obtained using the sensor(s).

The system is able to modulate biologically based recommendations based upon sensor output, and thereby nudge or encourage / discourage use of certain products. The degree of modulation may be tuneable by the user, i.e. to alter the effect of sensor data on the biologically based recommendations.

Also described herein is a computer implemented method comprising determining cut off values for a plurality of nutritional components based on an individual’s personal biological information, modulating or adjusting those cut-off values based on current or recent physiological or biochemical functions of the individual such as activity, and applying the modulated cut-off values to products, such as consumable or topically appliable products, to provide product recommendations.

Brief Description of the Drawings Figure 1 is a perspective schematic view of a wearable device according to an embodiment of the invention;

Figure 2 is a schematic system view of the wearable device of Figure 1 ;

Figure 3 is a flow diagram of the data processing performed by the wearable device of Figure 1 ;

Figures 4A and 4B are a flow diagram of the data processing performed by the activity classifier of Figure 3 (Figure 4B is a continuation of Figure 4A);

Figure 5 is a flow diagram of the data processing performed by the lifestyle classifier of Figure 3;

Figure 6 is a flow diagram of the data processing performed by the penalty state controller of Figure 3; and

Figure 7 is a flow diagram of a method of providing product recommendations to a user;

Figure 8 is a flow diagram of the data processing performed by the wearable device of Figure 1 ;

Figure 9 shows a flow diagram in which product recommendations are adjusted according to a user’s predicted calorie sensitivity and activity data of the user;

Figure 10 illustrates schematically the closed loop system for modulating product recommendations.

Figure 1 1 is a schematic view of two graphical user interface elements;

Figure 12 is a graph showing a relationship between computed sitting time and the percentage of products selected for modulation;

Figure 13 is a table showing a graphical user interface element which is updated over the course of a 24-hour period according to user activity;

Figure 14 is a schematic view of a graphical user interface incorporating one of the graphical user interface elements of Figure 1 1 ;

Figure 15 is a schematic view of a graphical user interface incorporating one of the graphical user interface elements of Figure 1 1.

Detailed Description

A user’s genetic profile (genes) can be analysed to determine his or her risk, or likelihood, of developing long-term chronic diseases, such as Obesity, Type 2 Diabetes Mellitus and Cardiovascular Disease. These genetic risks are fixed variables that cannot be adjusted. However, there are several adjustable factors that can reduce a user’s risk of developing chronic diseases, such as diet and physical activity.

The embodiments described here aim to address the problems described above by adapting the product recommendations described above using measurements indicative of the user’s physiological functions, such as measurements indicative of the user’s calorific expenditure during the last week or heart rate data of the user. By taking into account other (e.g. non-genetic) factors which contribute to chronic disease risk, a user is able to select products which are more likely to benefit his or her health.

For example, personalised food recommendations may be provided based on both an individual’s genetics and their physical activity levels, measured using a sensor such as an accelerometer. Personalised product recommendations may also be obtained for other classes of product such cosmetics, medicine, drugs, vitamins etc.

A genetic testing service (provided by DnaNudge, London, UK) provides personalised food recommendations to individuals based on their genetics. An individual undergoes a DNA (or RNA) test to be assessed on several Single Nucleotide Polymorphisms (SNPs). These SNPs have been identified in scientific literature, such as Genome Wide Association Studies (GWAS), as being associated with several chronic diseases, e.g. Obesity and Type 2 Diabetes. The results of the genetic test are categorised into five layers - Very Low Risk, Low Risk, Medium Risk, High Risk, and Very High Risk. The genetic results of an individual are then correlated to six nutrients: Calories, Fat, Saturated Fat, Carbohydrates, Sugar and Salt. From these correlations a set of nutritional cut-offs have been developed. These nutritional cut-offs form the basis for the personalised food recommendations. For example, if a product contains a level of salt which exceeds the nutritional cut-off value for salt then the product will not be recommended to the user.

Personalised“on-the-spot” food recommendations can, for example, be provided to a user using a wearable device, such as a wristband device (referred to as a“DnaBand”). The wearable device can also monitor the physical activity of the wearer and determine one or more physical activity factors which reflect the amount of physical activity the wearer has undertaken while wearing the device. A person’s physical activity level is factored into their baseline genetic recommendations. For example, the physical activity factors are combined with the nutritional cut-offs to update the personalised food recommendations. The recommendations are better targeted to the user because they take into account that both diet and physical activity influence chronic disease risk.

One type of physical activity factor is a“calorific cut-off” which is used to modulate a nutritional cut-off value for calories. For example, if a user is determined to have been relatively physically inactive over the past week or so, then a relatively low“calorific cut-off” value may be generated. If this calorific cut-off is below the nutritional cut-off value for calories (determined from the results of the genetic test), then the lower calorific cut-off value may be used to generate the product recommendations. For example, a user who has no genetic pre-disposition towards obesity may have a high nutritional cut-off value for calories. However, if the user has not done much exercise recently, this value may be lowered accordingly, such that a product which has a high calorific content, such as a packet of crisps, is recommended as being unsuitable for the user.

The wearable device therefore provides feedback on an individual’s physical activity data, adjusting the calorie cut-off of the individual. The calorie cut-off will be decreased if physical activity is inadequate and reintroduced to baseline when physical activity becomes adequate. This combination between physical activity, diet and genes creates a closed-loop feedback system providing more accurate personalised food recommendations. A user may be able to control the degree of feedback so as to vary the amount by which product recommendations are modulated based upon physical activity. For example, a user may not want product recommendations to be influenced in any way by physical activity in which case the modulation is set to zero. Another user may want this influence to be very high, in which case the modulation is set to high value. The user may control modulation using some means provided on the band, or via an interface with a computer device such as a smartphone.

Whilst a lack of physical activity has long been recognised as detrimental to health, the negative effects of sitting are generally underappreciated. Ekeland et a!., The Lancet 2016, 1303-1310 (388), identifies that sitting is critical to health and recommends that three times the recommended daily activity is required to compensate for extended periods of sitting. However, it is not always possible for users to undertake the recommended amount of daily activity, let alone three times the recommended daily activity. Configuring the wearable device so as to adjust product recommendations according to the device wearer’s sitting behaviour may go some way to mitigating for extended periods of sitting. Additionally, for many users, a clear and visible indication that certain products are no longer recommended because they have been sitting for too long, may in itself act as a motivator to avoid prolonged periods of sitting and to engage in physical activity.

The device and method presented here flow from a realisation that an individual’s health can be improved by using a wearable computer device to provide product recommendations that are modulated depending upon the individual’s sitting behaviour as detected by the device.

Figure 1 illustrates a wearable device 100 (or“wristband”) comprising a strap 101 , which in this example has an expandable section 102 to allow the user to slide the wristband 100 easily over his or her wrist. In other examples, a wrist strap, such as those found on wristwatches, may be used in place of the strap with expandable section. Figure 1 also shows three orthogonal axes X, Y and Z used to describe the orientation of the wristband 100. The Y axis is oriented in along the axis of the wristband, i.e. the direction in which the wearer’s wrist passes through the wristband 100. The X and Z axes are perpendicular to the Y axis (and to one another), with the Z-axis aligned from top to bottom of the wristband 100, i.e. the Z axis points from the outside to the inside of the wrist when the wristband 100 is worn with the expandable section 102 on the inside of the wrist.

The wristband 100 comprises an optical sensor 103, such as a photovoltaic cell or camera, and a light source 104, such as a laser. A window 105 is provided in the wristband 100 to allow the optical sensor 103 to be used for reading product codes and the light source 104 to be used to illuminate the product code. An indicator (or indicators), such as a light emitting diode (LED) 106, is also provided in order to give feedback to the wearer about a product. The cross section of the strap 101 is thicker on one side in order to accommodate the various components of the wearable device 100 (see below).

Figure 2 shows a schematic system view of wristband 100. The wristband 100 is powered by a battery 201 , which can be charged using a charging unit 202, and has inertial sensors for measuring the motion of the wristband 100 in 3D space, an accelerometer 203 (e.g. a tri-axial accelerometer) and/or a gyroscope 204. The axes of the accelerometer 203 and/or gyroscope 204 are aligned with respect to the three orthogonal axes X, Y and Z shown in Figure 1.

The wristband 100 may also comprise additional sensors and electrodes 205, such as a heart monitor (e.g. an electrocardiogram, ECG) or thermometer for measuring the user’s heart rate or temperature, and/or a GPS sensor (or other positioning system) for tracking the user’s location. Sensors may include, for example, a microphone or optical sensor for measuring heart rate.

The wristband 100 is controlled by a processing unit 206 which accesses instructions and data stored in a memory 207. A wireless communication module 208 is provided to allow the processing unit 206 to communicate with other computer devices such as other wristbands, smartphones, smartwatches or personal computers. The wireless communication module 208 can be used, for example, to provide or update a database of product codes and/or product recommendations stored in memory 207. The wireless communication module 208 may allow data to be exchanged between wristbands 100.

The wearer may activate the light source 104 and direct light onto a product in order to read a product code (or other information) from the product using the optical sensor 103. The wearer may manipulate or hold the product using either one of their hands, or both of their hands together, in order to orient the product correctly for reading. Alternatively, the wearer may move or orient the wristband 100 in order to read the product code whilst the product remains in place on a supermarket shelf (for example). After reading the product code, the wristband may provide feedback to the wearer using the indicator (LED) 109, which may be an RGB LED which can be adjusted to show different colours by mixing different combinations of the RGB components and/or adjusting the brightness and/or blinking pattern of the LED. Of course, other types of indicator can also be used, such as LED arrays or screens such as LCD, LED or OLED screens.

Examples of the sensors which may be incorporated into the wearable device are:

• Inertial sensors, such as an accelerometer (e.g. a tri-axial accelerometer) and/or gyroscope; • Pedometers/step counters;

• Pulse rate sensors, e.g. photoplethysmography (PPG) sensors;

• Respiration rate sensors;

• Heart rate sensors (also for measuring heart rate variability);

• Blood pressure sensors;

• Microneedles for performing in situ blood tests e.g. of blood glucose levels;

• Air quality or pollution sensors (e.g. mass spectrometers);

• UV light monitors (e.g. photo-diodes).

An indicator 209, such as a light emitting diode (LED) and/or a vibrator 210, is provided in order to provide visual or tactile feedback to the user wearing the wristband 200. In one example, the indicator 209 may provide the product recommendations according to a“traffic light system”, with a“red” colour indicating that a product is not recommended for the user or a“green” colour indicating that a product is recommended for the user. A further“amber” colour may also be used to indicate that a product would have been recommended as suitable for the user had the product recommendation not been adjusted based on the user’s activity (e.g sitting or other physiological function). It should be understood that the references to colours here is not intended to be limiting and other ways of indicating a binary or a three-level (or a large number of levels) recommendation system may be used. For example, the wristband 100 may display a representation of a product’s nutritional information together with the nutritional cut-off values determined for the user and/or the modified cut-off values.

Various steps in controlling how the wearable device 100 provides product recommendations to the wearer will now be described. In particular, various steps in generating updated product recommendations for the exemplary cases of sedentary behviour monitoring or physical activity monitoring are described. It will be appreciated, however, that the techniques described below can be applied to other types of monitoring relating to one or more physiological and or biochemical functions of the user, or of a user environment.

Sedentary behaviour

Figure 3 shows how the motion data 301 collected by the inertial sensors 203, 204 is processed to update the product recommendations. In this example, the motion data 301 includes acceleration components (x, y, and z) along three orthogonal directions of the wearable device 100 in 3D space, measured at a series of time steps (x) by the tri- axial accelerometer 203. The motion data 301 may also include data derived from the acceleration components such as the position or velocity of the wearable device 100. The motion data 301 may also include the orientation (or the angular velocity or angular acceleration) of the device in 3D space, as determined by the gyroscope 204, for example.

The motion data 301 is provided to an activity classifier 302, typically at a sampling rate of 40 Hz. As described in more detail below with reference to Figures 4A and 4B, the activity classifier 302 processes the motion data 301 to determine an activity class for the wearer of the device, such as sitting, inactive, active or unknown. The activity class can be updated when each sample of the motion data 301 is received or, more commonly, after a threshold number of samples of motion data 301 is received, such that the activity class is updated with a lower frequency than the sampling rate.

The activity classifier 302 provides activity class data 303 to a lifestyle classifier 304, typically at a sampling rate of 1 Hz. As described in more detail below with reference to Figure 5, the lifestyle classifier 304 accumulates the activity class data 303 and determines from the data a list of accrued penalties 305 according to the behaviour of the device wearer. Penalties are accumulated over a period of 24 hours, for example.

The lifestyle classifier 304 provides the list of penalties 305 to a penalty state controller 306, typically on a daily basis, although a shorter interval (e.g. hourly) or a longer interval (e.g. weekly) could also be used. As described in more detail below with reference to Figure 6, the penalty state classifier 306 determines a penalty state for the wearable device 100 based on the list of penalties 305. The penalty state may also be based on other factors, such as information about the wearer, e.g. their age, sex or product purchase history (which in some embodiments is measured using the wearable device 100). When the penalty state controller 306 determines that the device 100 should be placed in a penalty state, it sets a flag in the memory 207 to indicate that the device 100 should provide adjusted product recommendations. If no penalty state is determined then the flag may be unset such that the device 100 provides product recommendations which are not affected by the wearer’s sitting behaviour. Figure 4A shows how the motion data 301 obtained from the inertial sensors (referred to in this figure a Motion Processing Unit, MPU) is processed by the activity classifier 302. Once the wearable device 100 is turned on, the activity classifier 302 begins to receive samples 401 from the MPU. Each time a sample is received, the magnitude (e.g. Euclidean norm) of the acceleration is calculated from the x, y and z components of the acceleration measured by the accelerometer 203. After a threshold number of samples have been received 403 (in this case, 40 samples), the variance of the acceleration is calculated 404 for the samples. The variance is then compared to a threshold 405 to determine whether or not the device wearer is relatively static. If the threshold is exceeded then a further comparison is made to determine if the variance exceeds another threshold associated with the variance expected for slowly walking. If this latter threshold is exceeded then the activity type is classified as“High” 407; if not, the activity type is classified as“Medium”.

Returning to the comparison of the variance of the acceleration with the threshold for static behaviour 405, if the variance is less than this threshold, the sample data is filtered and processed 409 to determine whether or not the device wearer is sitting. The filtering removes unwanted noise from each x, y and z component of the acceleration, e.g. using a band-pass filter. The filtered acceleration components can be used in a number of ways to determine whether the wearer of the device is sitting. For example, the contribution of the acceleration due to gravity in the filtered acceleration components can be used to determine the lower arm posture of the wristband wearer (assuming that the wristband is worn in a conventional way), and hence whether the wearer is performing a low level of activity while standing or sitting. In the exemplary formula shown in element 409 of the figure, the filtered components are compared with each other to identify whether the acceleration of the device 100 along one of the axes (in this case the X direction) is greater than the acceleration along each of the two other axes, as this is typically correlated with the wearer performing a low level of activity while standing/not sitting, e.g. the wearer swinging his or her arms backwards and forwards while walking. The details of how such a comparison is implemented depend on the orientation of the accelerometer 203. A gyroscope 204 may be used in conjunction with the accelerometer 203 to improve sitting detection accuracy. Other methods for detecting sitting which can be used include machine learning methods in which a classifier is trained based on motion data which has been labelled according to the type of activity performed by the wearer of the device as a function of time. As a result of this analysis, the activity type for the period covered by the samples of motion data is determined to be either sitting 410 or low 41 1.

After the activity type has been determined, one of three counters is incremented depending on the activity type. If the activity type is“high” or“medium” then a counter for Moderate to Vigorous Physical Activity (MVPA) is incremented 412, whilst if the activity type is“sitting” or“low” then respective counters for sitting 413 and low 414 may be incremented. Regardless of the activity type, a counter for a prediction index is incremented 415 in order to count the number of times the activity type has been predicted. If the prediction index does not exceed a threshold, in this case 60 iterations, the activity classifier 302 waits for further samples to be received from the inertial sensors (MPU) 401 and the process is repeated. The threshold is chosen such that a reliable prediction of the wearer’s behaviour can be obtained.

Figure 4B is a continuation of Figure 4A and shows the process performed by the activity classifier 302 after the prediction index has been found to exceed the threshold, i.e. after a sufficient number of activity type classifications have been performed. A determination is made of whether the MVPA counter is above a certain value (in this case, 30 counts) 417 and, if so, the lifestyle type associated with the samples of motion data 301 is set as“active” 418. Otherwise, a determination is made of whether the low counter is above a certain value (in this case, 30) 419 is made and, if so, the lifestyle type is set as“inactive” 420. Otherwise, a determination is made of whether the sit counter is above a certain value (in this case, 30) 421 : if it is, then the lifestyle type is set as“sitting”, or if not, the lifestyle type is set as“unknown”. In other words, the final stages of the process 417-423 determine the predominant behaviour of the device wearer over the relevant period, i.e. whether the user has been active or sitting for more than half the time. After this has been done, the counters are each reset and the process of Figures 4A and 4B is repeated.

Figure 5 shows how the lifestyle classifier 304 processes the lifestyle types obtained from the activity classifier 302. The lifestyle classifier 304 has counters for activity (AC), inactivity (IC) or sitting (SC) which are initially set to zero 501. The initial state of the classifier 304 is set to be“NOT SITTING”. On receipt of a lifestyle type from the activity classifier 302 (referred to here as“Level 1”) 503, the lifestyle classifier 304 determines the type and increments the respective counter, whilst setting the other counters to zero 505, 506 and 507. For example, when the activity counter is incremented then the inactivity and sitting counters are set to zero 505. After incrementing the activity counter or the inactivity counter the state of the classifier 304 remains“NOT SITTING” and the process 502-504 is repeated for the next lifestyle type determined from the motion data 301. If the sitting counter is incremented 507 then a determination of whether the sitting counter has reached a particular value 508 (in this case, 2 counts). In other words, steps 504 to 508 determine whether the device wearer has been sitting continuously for a certain period (2 cycles), i.e. without their sitting being interrupted by a period of MVPA or low-level non-sitting activity. If this criterion 508 has not been met, the state of the classifier 302 remains“NOT SITTING” and the process is repeated again. However, if the criterion 508 has been met, then the activity and inactivity counters are set to zero 509 and the state of the classifier 304 is set to “SITTING”.

Once the lifestyle classifier 304 enters the“SITTING” state 510, it continues to receive lifestyle types from the activity classifier 302. The lifestyle classifier 304 processes the lifestyle types 51 1 , 513 in a similar way as when it is in the“NOT SITTING” state. However, after either of the activity 513 and inactivity 514 counters is incremented (and the other counters set to zero) a determination is made of whether the either of the activity and inactivity counters exceeds a certain value (in this case, 2 counts). In other words, the classifier 304 determines whether the device wearer has undertaken MVPA continuously for the last two cycles or low-level non-sitting activity continuously for the last two cycles. If they have, then the activity and inactivity counters are reset to zero 516 and the state of the classifier 304 is returned to“NOT SITTING” 502. If not, then the sitting counter is incremented 517.

After the sitting counter is incremented, either at step 517 or 518, a determination is made of whether the sitting count exceeds a threshold 520 (in this case, 30 counts). If it does, a penalty is added to the penalty list 520 and the sitting counter is set to zero 521. The process then restarts 510 with the state of the classifier 304 remaining as “SITTING”.

Figure 6 shows how the penalty state controller 306 uses the penalty list to control the behaviour of the wearable device 100. When the device 100 is turned on or reset 601 , the controller 306 determines whether a certain period (in this case, 24 hours) has elapsed since the wearable device was last operational. If it has, then the state of the controller 306 is set to“GREEN” and a penalty score is also set to zero 604. In the “GREEN” state, the product recommendations are not modulated when the device 100 is used to read a product code, but rather the“normal” product recommendations (e.g. product recommendations based on the genotype of the user) are indicated to the user.

The controller 306 waits 605 to receive a new penalty from the lifestyle classifier 304. When a penalty is received, the penalty score is incremented 607 and a determination is made 608 of whether so-called“sedentary” rules have been met. These rules may differ according to user information, such as whether the user is a child or adult. For example, a rule for an adult may state that if there a penalty score of 12 or more in a 16 hour period then the user is classified as“sedentary”, whilst for the corresponding rule for a child may require a penalty score of only 6 or more. If the sedentary rules are not met 609, then a check is made whether a new day has started 609 (or some other period has elapsed) and, if not, the controller 306 continues to wait 605 for a further penalty point to be received from the lifestyle classifier 304. If a new day has started, the state of the controller remains“GREEN” 610 and the penalty score is set to zero 604, as previously.

A failure to meet the sedentary rules (or rule) causes the state of the controller 306 to become“AMBER” 61 1 . In this state, the product recommendations provided by the device 100 are modulated. The controller 306 remains in this state until the next day 612, after which the penalty score is set to zero once more 604 and the state is set back to“GREEN” (not shown) .

In one example, the product recommendations may of three different types: not recommended, maybe recommended and recommended. These types may be indicated to the user using the colour-changing LED 106 of the device 100. For example, a traffic light system of red (not recommended), amber (maybe recommended) and green (recommended) can be used. When the controller 306 is in the“GREEN” state, then“normal” product recommendations are provided to the user. However, if too many penalty points have been accrued and the controller 306 enters the“AMBER” state, a subset of the product recommendations which would otherwise have been indicated using the green light may be indicated to the user with an amber light. For example, a food product with a moderately high calorie content may usually be recommended as being suitable for a particular user (i.e. indicated with a green light). However, if the user has been sitting for a prolonged period, the product recommendation may be adjusted (modulated) to“maybe recommended” (i.e. indicated with an amber light).

The controller 306 can also be configured so that a subset of products which are normally “maybe recommended” (i.e. indicated with an amber light) become “not recommended” (i.e. indicated with a red light) based on the sitting behaviour of the device wearer, as described above. However, products which are“not recommended” are, in general, not affected by sitting behaviour (i.e. always indicated with a red light).

The use of three“states” of product recommendation described above is provided by way of example only. It is of course possible to use only two product recommendation states (“recommended” or“not recommended”) or to use more than three product recommendation states, e.g. a recommendation score, which may be out of 100, for example or may be positive or negative, with negative values indicating that a product is not recommended and vice versa. Where a recommendation score is used, the modulation may be applied by decreasing the score based on the number of penalty points accrued by the wearer.

The subset of products for which the recommendation is modulated by sitting behaviour may be chosen in a number of ways. In general, the modulation is not applied to all types of food product because to do so may cause“healthy” products such as vegetables to be“maybe recommended” based on the user’s sitting behaviour. The subset may be chosen based on nutritional data associated with each product. In one example, the products which are classified as being“recommended” are ranked according to their calorific content and the top 50%, 30% or 10% of the products with the highest calorific content are selected for modulation. In some embodiments, the cut-off percentage for selecting the products this way can be adjusted by the user in order to increase or decrease the number of products for which the modulation is applied.

In general, the following procedure may be used to obtaining product recommendations. Firstly, a number of health traits associated with potentially poor health are identified, including for example, type-ll diabetes, hypertension, high body mass index (BMI), and high cholesterol. Whether or not a user has any of the these health traits may be determined using genetic testing e.g. based on detecting single nucleotide polymorphisms (SNPs) in a sample of the user’s DNA, although other forms of diagnostic testing can also be employed, such as testing the user’s breath to determine the composition of their microbiome.

Once the“target” health traits have been identified, the effect of diet on each of them is considered. This is done by analysing how the health trait is likely to be affected for each of a number of different categories relating to the nutritional content of consumable products, including any of the following, for example: sugar content, calorific content, carbohydrate content, saturated fat content, total fat content, and salt content. For example, it is known that the risk of type-ll diabetes can be raised by consumption of products which are high in sugar, whilst salt-rich diets are linked to hypertension. The relationship between each category and each health trait can be quantified by assigning a trait-dependent coefficient to each of the categories. Taking hypertension as an example, large coefficients may be assigned to salt content and fat content, whilst small (or even zero) coefficients may be assigned to calorific content and carbohydrate content.

For each identified health trait, the trait-dependent coefficients can be used to calculate a score which indicates the expected effect of a product based on its nutritional content, i.e. how many grams of salt, saturated fat, etc the product contains. For example, to calculate a score which is indicative of the adverse effect of a product on hypertension, the number of grams of salt in the product can be multiplied by the coefficient for salt content. Mathematically speaking, the score for each trait may be determined by taking a scalar product of a vector of the trait-dependent coefficients with a vector of nutritional information for each product. Of course, more sophisticated ways of calculating the scores can also be used, such as using a dose-response curve determined for a health trait to model the likely effect of the different types of nutritional category on that trait. The score may also take into account other factors, such as the type of product (e.g. confectionary, biscuits, breakfast cereals etc.) and the typical portion size for the products. These other factors may be used to adjust the scores for differences in how different products are consumed, e.g. if a product is identified as typically being consumed as a“treat” then its score may be reduced so that its recommendation may be more positive than it would otherwise have been.

After the scores have been calculated for each product, the products are ranked in descending order according to their score for a particular health trait. A subset of the products is then chosen by selecting products which have a ranking which is higher than a threshold ranking, e.g. products which are in the top 50% or 30% or 10% of products based on their score of the health trait. These products may be assigned a “negative” recommendation, such as“not recommended”, so that when the wristband 100 is used to read the barcode of the product, the LED 104 turns red (for example). The remaining products are assigned to a different category, e.g. “recommended” (which would be indicated to the user with a green colour). Thus, the products are allocated different recommendations according to their ranking for a particular trait. The value of the threshold ranking or “cut off” which is used for allocating the recommendations varies from user to user and is determined based on how susceptible the user is to the particular trait (as determined by genetic testing, for example).

The potential effects of sitting (or other forms of sedentary behaviour) are accounted for by selecting a further subset of products for which the associated product recommendations are modulated depending on the penalty state of the controller 306. This subset may be determined by selecting products which are below the threshold ranking mentioned above, but above a second threshold ranking. For example, if 50% of the products are“recommended” for a particular trait, then the top 20% of these products may be chosen for the subset which is to be modulated. The second threshold ranking is chosen based on how severely the particular health trait is affected by sitting and/or other types of sedentary behaviour. The subset of products may be assigned to a particular database table or otherwise“flagged” in the memory of the wristband 100 so that it can be easily determined whether the product recommendation for a particular product should be modulated. In one example, if the controller 306 is in the “AMBER” state, the products in the subset are indicated as being “maybe recommended” (e.g. the indicator turns amber) whereas if the controller 306 is in the “GREEN” state, they are“recommended” (e.g. the indicator turns green). The product recommendations for each health trait are combined to determine an overall product recommendation for each product. This can be done in a number of ways, such as determining that the product is“not recommended” if the product is“not recommended” for any of the traits or, alternatively, if the product is “not recommended” for more than one of the traits. The flag indicating whether the product recommendations should be modulated can also be set or unset for each product based on the same or similar kinds of rule.

In general, the threshold rankings mentioned above, and the rules for determining the overall product recommendations, strike a balance between allowing users to choose products they wish to consume and dissuading them from consuming at least some of the products which are most likely to have a negative impact on their health. This balance is advantageous because the user is“nudged” into making choices which are better for him or her in the long term. Modulating the product recommendations based on the user’s sitting and/or other sedentary behaviour provides an additional“nudge” so that, for example, a user may notice that a product which is normally “recommended” is only “maybe recommended” (or recommended as being “discretionary”), if the user has been sitting for long periods during a particular day. Using a modulation which does not make the products“not recommended” may be particularly advantageous because it does not penalise the user too harshly for circumstances which may be out of their control, e.g. if the user has been sitting for a long time because they have had to travel a large distance.

Figure 7 provides an overview of the steps involved in a method of providing a user with product recommendations. The steps of the method are as follows:

Step 701 : Determine one or more health traits of the user, such as: type-ll diabetes, hypertension, high body mass index (BMI), and high cholesterol, e.g. by performing a genetic test on a biological sample provided by the user.

Step 702: For each of a multiplicity of consumable products, calculate a score indicative of how the product affects each of said health traits, each score being based at least partly on nutritional information for the product. The scores may be calculated using the procedure described above, e.g. for each product, determining its nutritional content each of a number of categories (e.g. using product nutritional information provided by the manufacturer of the product), multiplying the nutritional content in each category by a trait-dependent coefficient and summing the resulting values to produce an overall score for each trait.

Step 703: Assign a product recommendation to each product based on the scores. As described above, this may be done by ranking the products according to their respective scores for each of the health traits, selecting subsets of the ranked products for each health trait, assigning a recommendation to the products in each subset and then combining the recommendations for each health trait to produce an overall product recommendation for each product. The product recommendations may be chosen from a pre-defined set of product recommendations.

Step 704: Monitor the behaviour of the user by identifying periods when the user is in a sitting position or other sedentary state.

Step 705: Based on the behaviour of the user, modulate the product recommendation for at least of a subset of the products. The product recommendations may be modulated, for example, by assigning a different product recommendation from a pre defined set of product recommendations. The subset may be determined based on the the products’ scores for each health trait, e.g. a certain percentage of products which were“recommended” (i.e.“green”) for a particular trait may be re-assigned to“maybe recommended” (i.e.“amber”).

Step 706: Provide the modulated product recommendation to the user via a visual indicator, e.g. the light source 104 of the wristband 100.

Physical activity and environmental monitoring

Figure 8 shows a block diagram of how signals from an inertial sensor 801 in the wearable device are processed according to one implementation. In this case, the sensor provides signal data associated with motion around three orthogonal axes (x, y and z). These signals are sampled (together with signals from any other sensors in the wearable device). Signals from the inertial sensor(s) are sampled at least at 20 Hz (i.e. a Nyquist frequency of 10 Hz), in order to capture all the signal content related to moderate and high intensity physical activity/exercise (e.g. walking and running). The sampled (i.e. raw) signals are then pre-processed 802 by filtering. For inertial signals, a band-pass (BP) filtering pipeline is implemented with a bandwidth of 0.25 to 8 Hz to remove unwanted noise components whilst ensuring all components of the signals related to moderate and high intensity activities remain. Second order Butterworth filters are used as they provide a smooth transition between pass and stop bands, as well as a uniform unit gain at the pass band. The filters are designed using zero-pole analysis to ensure their stability. Non-linear effects of the phase response were removed by applying a zero-phase filtering technique in which the signal is filtered forward and then backwards.

The signals are analysed in segments of configurable length, although typically with lengths (durations) in the range from 1 s to 60 s. For each segment or“window” of the pre-processed accelerometer data, an average magnitude or“energy” is calculated 803. These average values may be termed“Activity Accelerometer Counts” (AAC). For example, if a tri-axial accelerometer is used, the (typically rectified) components of the acceleration measured along each of the axes may be summed and an average calculated over the different (discrete) times within the window for which the data has been obtained, e.g. using a numerical quadrature rule such as Simpson’s rule. Alternatively, the vector norm (i.e. 2-norm) of the components can be used to calculate a total acceleration which is averaged

A physical activity (PA) classifier 805 is then used to determine whether the user has been physically active for each time window. For example, this can be done by applying a simple threshold rule to the windowed AAC data, i.e. if the AAC value for a particular window exceeds a specified value then the user is determined to have been physically active during that period. A suitable threshold may be determined by, for example, by measuring the AAC values obtained from a user performing different types of physical exercise at different levels of intensity. More sophisticated classifiers may be used to determine the intensity or type of physical activity undertaken, e.g. to distinguish between moderate or very high levels of activity or between running and cycling.

The classifications of whether the user has been physically active or not are then used to place the person into a binary category - either ‘Inactive’ or ‘Active’. What determines if someone is classified as‘Inactive’ or Active is if the physical activity classification values meet evidence-based guidelines regarding physical activity, such as those provided by the National Institute for Health and Clinical Excellence (NICE, a United Kingdom government organisation). These guidelines outline the expected amount of physical activity for various age groups.

For example, the NICE Physical Activity Guidelines (PAGs) for 19-64 years old, recommend:

• Adults should aim to be active daily. Over a week, activity should add up to at least 150 minutes (2½ hours) of moderate intensity activity in bouts of 10 minutes or more - one way to approach this is to do 30 minutes on at least 5 days a week.

• Alternatively, comparable benefits can be achieved through 75 minutes of vigorous intensity activity spread across the week or combinations of moderate and vigorous intensity activity.

• Adults should also undertake physical activity to improve muscle strength on at least two days a week.

• All adults should minimise the amount of time spent being sedentary (sitting) for extended periods.

In one implementation of the NICE guidelines, if an individual does less than 150 minutes moderate intensity physical activity per week or less than 75 minutes vigorous intensity or combined moderate and vigorous intensity physical activity per week then the individual is ‘Inactive’. If the individual meets this requirement they can be considered‘Active’ based on guidelines.

The physical activity guidelines are based on a weekly guideline; therefore, the closed- loop feedback system is dynamic and personalised food recommendations will be changed based on whether a user has become‘Active’ or‘Inactive’ (and vice versa). Changes in the average level of user activity can be taken into account by calculating a moving average of the activity data, e.g. by performing the categorisation of whether a user has been active or not over the past week. Of course, other averaging times can be used, such as 1 day, or about 1 month.

The resultant category (i.e. in this example, active of inactive) is then passed onto a decision stage 805 that uses this and other information (DNA and/or nutritional information) to determine the recommendation updates (e.g. green to amber) for the products loaded in the band memory. Alternatively,“modifier” values can be stored in the device (or remotely) and applied“on the fly” to update a recommendation after the user has scanned a product.

As discussed above, the physical activity category may be used to adjust the calorie cut-off value which is used to determine whether a product is recommended or not based on its calorific content. This adjustment is dependent on whether an individual has been categorised into either‘Active’ or‘Inactive’.

• Active - If a person meets the PAGs then personalised food recommendation remain solely based on genetics and there is no further tailoring of recommendations

• Inactive - If a person does not meet the PAGs then the nutritional cut-off for calories only will be tailored. A person’s calorie sensitivity will be increased, therefore reducing calorie cut-off and subsequently calorie allowance.

Energy balance is one of the key factors regarding weight management. Energy can be measured in either calories or kilojoules and is derived from the total amount of protein, fat and carbohydrate in foods. The key to long term weight management is ensuring the correct balance between the number of calories an individual consumes (input) and the number of calories that is utilised (output).

Three scenarios for weight management can be identified according to the level of energy balance: (i) if calorie intake is greater than total energy expenditure, weight gain will occur; (ii) if calorie intake equals total energy expenditure, a constant weight will be maintained; and (iii) if calorie intake is lower than total energy expenditure then weight loss will occur. Therefore, to prevent the state of weight gain (as a result of a net calorific intake), an individual’s caloric intake needs to be tailored, i.e. decreased from baseline or increased to baseline.

This tailoring of the calorie cut-off will, in general, cause a number of products to go from a‘green’ recommendation to an‘amber’ recommendation. In this case, the colour amber indicates a food product is not recommended due to inadequate physical activity, and that had physical activity been adequate the product would have been a green recommendation. It is important, however, not to reduce the number of healthy foods recommended to people e.g. vegetables. Therefore, the calorie cut-off adjustment is only applicable to certain food groups, such as foods which are classified as“discretionary” for the user such as crisps, chocolates, sweets.

Various scenarios are now described in which product recommendations are adjusted according to different combinations of a genetically determined sensitivity or tendency of the user and measurement data.

Figure 9 illustrates the process of adjusting product recommendations according to a user’s predicted calorie sensitivity and the activity data of the user. The process comprises the following steps:

A: Products: calorie content

B: DNA Result: calorie sensitivity result (all users)

C: Real time measurement: activity: steps/distance within given time and as compared to baseline.

D: Recommendations adjusted: high calorie product not recommended.

E: Pop up: to encourage more users to be active.

The following examples are similar to the example described with reference to Figure 9, except that the steps A-E are replaced with the steps provided for each example.

In one example, the product recommendations are adjusted according to a user’s predicted rate of metabolising caffeine and the time of day. Caffeine has a longer lasting effect on“slow” caffeine metabolisers. Continuous real time measurements of the time of day and/or the user’s heart rate may be used to adjust whether a particular caffeine containing product, such as coffee or an energy drink, is recommended. A: Products: coffee, tea, sports drinks, protein shakes, fizzy drinks (coke).

B: DNA Result: slow metaboliser Caffeine SNP rs762551 .

C: Real time measurement: Time of day

D: Recommendations adjusted product not recommended

E: Pop up: to remind users of their DNA result

In another example, the product recommendations are adjusted according to a user’s predicted susceptibility for hypertension and heart rate data of the user. For users with high resting heart rate, the original (i.e. “healthy”) fat recommendations within categories can be adjusted to encourage consumption of e.g. oily fish & nuts. Similarly, the nutritional cut off value for salt can be reduced and/or supplements can be recommended, e.g. omega 3, 6, 24. The cut off values based on heart rate can be further adjusted over time.

A: Products: total fat - but product category important, product salt and saturated fat content.

B: DNA Result: hypertension SNPs.

C: Real time measurement: heart rate

D: Recommendations adjusted: category dependent - encourage healthy fats, decrease saturated fat content.

E: Pop up: to remind users to avoid saturated fat/salt for heart health and encourage users when scanning heart-healthy product.

In another example, the product recommendations are adjusted according to measurements of a user’s sweat production. Vitamin recommendations can be adjusted dependent on sweat level.

A: Products: isotonic solutions, sports supplements such as protein shakes

B: DNA Result: all users

C: Real time measurement: sweat production

D: Recommendations adjusted: encourage isotonic solutions/protein supplement recommendations.

E: Pop up: to encourage supplement if necessary. In a further example, the product recommendations are adjusted according to a user’s predicted sun sensitivity and measurements of the user’s exposure to UV light. The exposure level may be determined by tracking user location and using a UV reference map to understand how exposed users are to UV. This approach can also be used to determine the levels of pollution to which the user has been exposed. UV sensors may also or alternatively be used, e.g. by integrating a photodiode in the wearable device. This information can be used to change SPF recommendation, e.g. so that high protection sun cream is recommended over lower SPF sun cream.

A: Products: with/without SPF (Sun Protection Factor)

B: DNA Result: examples of genes that relate to sun sensitivity NTM AA, TYR GG,

ASIP TC, LOC10537 CC.

C: Real time measurement: LED, UV/VIS spectroscopy

D: Recommendations adjusted change SPF cut off.

E: Pop up: alert if users have had high UV exposure

As another example, the product recommendations are adjusted according to a user’s predicted ability to produce vitamin E and measurements of the user’s exposure to UV light. UV light (and sun exposure) reduces vitamin E levels in skin. Vitamin E can absorb the energy from ultraviolet (UV) light. UV maps (location based) or inbuilt UV measurements can be used to change user’s product recommendations to favour Vitamin E promoting ingredients

A: Products: with/without Vitamin E

B: DNA Result: Vitamin E SNP

C: Real time measurement: LED, UV/VIS spectroscopy

D: Recommendations adjusted: change Vitamin E cut off.

E: Pop up: alert if users have had high UV exposure.

In a yet further example, the product recommendations are adjusted according to a user’s predicted likelihood of suffering collagen degradation and measurements of the user’s hydration level and/or skin oil levels (using a sebumeter) and/or skin pH. Products may be recommended (or not) based on their oil content and/or pH balance.

A: Products: all skincare products B: DNA Result: genes for poor collagen degradation (dehydration can enhance appearance of wrinkles)

C: Real time measurement: hydration level as measured by e.g. Corneometer

D: Recommendations adjusted: encourage use of products containing humectants, occulsives and emollients

E: Pop up: to encourage users to keep hydrated for optimal skin health.

In a further example, the product recommendations are adjusted according to a user’s predicted likelihood of being adversely affected by pollution and measurements of the user’s exposure to pollution. The rationale for this is that pollution causes skin damage. The NQ01 gene influences a person’s ability to tolerate environmental toxins. There is a growing awareness of the negative impact of PM 2.5 - fine particulate matter, an airborne mix of tiny solid particles and liquid droplets, particularly its effect on city-dwelling consumers. Cosmetic users are concerned about pollution and ‘Anti-pollution’ is a new cosmetics industry. These types of product may advantageously be recommended to users who might be expected to have a tolerance to environmental toxins but have been exposed to very high levels of pollution.

A: Products: skincare products with/without anti-pollution ingredients

B: DNA Result: NQ01 gene (normal/poor pollution defence capacity)

C: Real time measurement: Location detection and look-up in a pollution map based on location, or test for particulates, e.g. fine particulate matter (PM 25 ).

D: Recommendations adjusted: promote anti-pollution recommendations.

E: Pop up: to remind users of their DNA result/tell user they have been exposed to pollution.

Figure 10 illustrates schematically the closed-loop approach to providing product recommendations. Personalised genetic (or other biologically derived) data 1001 is stored in the database 1003. This is used, as described, to generate cut-off values or thresholds for different nutritional components, e.g. carbohydrates, fat, salt, etc. These values are modulated, up or down at modulator 1005, based upon physiological and/or biochemical (or environmental) functions determined by a unit 101 1 that receives sensor data from the wearable device 1007. Using the modulated cut-off values, and product data, the wearable 1007 provides product recommendations 1009. Of course, all of the components illustrated in the Figure may be provided within the wearable 1007.

Alerting users to the effects of their behaviour and/or environment

An individual can be motivated or“nudged” to improve his or her lifestyle by providing them with more effective feedback of the consequences of sedentary behaviour. In particular, providing users with a visual indication of the number of product recommendations that have been adjusted because of their sedentary behaviour helps alert them to the fact that their diet (for example) should be changed to compensate for their sedentary behaviour. Users may therefore be motivated to avoid prolonged periods of sitting and to engage in physical activity to reduce the number of products for which the product recommendations that are adjusted.

Once the periods when the user has been sitting in a sitting position have been identified, e.g. according to the methods described above, the periods/sensor data may be analysed to determine how the product recommendations should be modulated. For example, if the user has sat for 30 minutes (for example), then the user may be awarded a“sedentary point”. In general, the sitting time is accumulated so that the total time the user has spent sitting is used to determine the number of sitting points. Alternatively, in some implementations, each sedentary point may only be awarded to the user if the he or she has sat for longer than a pre-determined period without a break of more than 2 minutes (or some other short period). The products for which the product recommendations are modulated may then be determined from the number of sedentary points. For example, the products may be ranked according to a score representative of the expected adverse effect of the product on a particular health trait or traits, e.g. how many grams of salt (or saturated fat, and so forth), that the product contains and the expected effect of this amount of salt on hypertension (for example). In this case, each sedentary point may correspond to an additional 10% of the ranked products having their product recommendation modulated or“downgraded”.

Another type of metric for sedentary behaviour is referred to as“Computed Sitting Time” (CST), which increases based on the amount of time the user has been sedentary but decreases in response to the user performing physical activity. For example, one version of this metric can be expressed mathematically as: CST = (Total number of TBLOCK - Total Steps / RSTEP) / max(T BL ocK).

In this equation, TBLOCK refers to the number of sedentary points, i.e. the amount of time for which the user has been sedentary (e.g. sitting) for a duration in excess of some threshold. In this example, the amount of physical activity undertaken by the user is quantified by the number of steps taken by the user divided by a constant factor, RSTEP, which defines the number of steps (e.g. 1000 steps) needed to cancel out one sedentary point (TBLOCK) . The CST may be calculated hourly. In this particular example, a factor“max(T BL oc K )” is used to ensure that the maximum CST is 1 per hour, e.g. max(T BL oc K ) is 2 when TBLOCK is calculated in half-hour intervals. In some implementations the value of TBLOCK may be set by the user.

The processing of the sensor data described above may be carried out by the wearable device 100. Alternatively or additionally, some or all of the sensor data produced by sensors of the wearable device 100 may be transmitted to a personal computer device (such as a smartphone) for processing, e.g. a personal computer device may determine when the user is sitting from sensor data provided by the wearable device 100.

Figure 1 1 shows an example of part of a graphical user interface (GUI), which includes a graphical user interface (GUI) element 1 101 that provides an overview of how the product recommendations for a particular user are distributed according to the various categories of product recommendations. In general, the product recommendations may be assigned to the various categories based on the user’s genotype, in which case the graphical user interface element 1 101 may be termed a“DNA Bar” or“DNA Products Bar”. The user interface element 1 101 may, for example, be displayed by the wearable device 100 and/or by a personal computing device (such as a smartphone).

In this example (and as described above), the product recommendations may of three different types:“not recommended”,“maybe recommended” and“recommended”. The user interface element 1 101 is divided into two“bars” or sections with each section corresponding to the one of the categories of product recommendation. The first section 1 102 has a length proportional to the number of products that are “recommended”. The second section 1 104 has a length that is proportional to the number of products that are“not recommended”. In general, the relative lengths of the first and second sections 1 102, 1 104 are fixed for a particular user, i.e. the subset of sedentary dependent products and the proportion of products that are always “recommended” or “not recommended” does not change according to the user’s behaviour. GUI element 1 101 therefore provides each user with a way of identifying their individual baseline or “starting point” in terms of the different product recommendation categories.

Another subset of products may be referred to as“Sedentary Dependent Products” (SDPs). Product recommendations for SDPs are modulated depending on the sedentary behaviour of the user. A further GUI element 305 is used to assist the user in monitoring how many of the SDPs have been modulated. The GUI element 1 105 is updated according to the sitting or sedentary behaviour of the user and may therefore be conveniently termed a“Health Bar”, or since it may be used in conjunction with the “DNA Bar” it may also be termed a“DNA Health Bar” or“Green DNA Bar”. These names serve to emphasise that the GUI element 305 may also help users to realise how their lifestyle reverses or“degrades” some of the advantages afforded to them by their genetic makeup (DNA). For example, users who are not genetically disposed to be at risk of obesity, but who have a generally inactive lifestyle, may be presented with product recommendations that are more characteristic of users who do have such a genetic predisposition. The DNA Health Bar provides a convenient and effective way for users to understand and be reminded how their lifestyle might be preventing them from“making the most” of their DNA.

In some implementations, the GUI element 1 105 can be integrated into graphical user element 1 101 , e.g. between the first and second sections 1 102, 1 104. The DNA Products Bar 1 101 may be updated relatively infrequently, such as when new products are added to the database or the product recommendations are re-computed based on new biological data provided by the user. By contrast, the DNA Health Bar 1 105 is typically updated multiple times in the course of a day in order to provide an dynamic indication of the proportion of products for which the associated recommendations which have been affected by the user’s sedentary behaviour.

The GUI element 1 105 is divided into a first subsection or“bar” 1 103A with a length proportional to the number of sedentary dependent products which are currently “recommended” based on the user’s behaviour, and a second subsection or“bar” 1 103B with a length proportional to the number of sedentary dependent products which are only currently“maybe recommended” based on the user’s behaviour. The visual style (e.g. colour or shading) of the first bar 1 103A may be chosen to match the first section 1 102 (to indicate that it represents products in the same product recommendation category), whilst the second bar may have a different visual style (to indicate that it represents products in a different product recommendation category). The first section 1 102 and the first bar 1 103A may both be green, whilst the second bar 1 103B may be amber, for example.

The relative lengths of the first and second bars 1 103A, 1 103B are indicative of the relative proportion of the SDPs for which the product recommendation has been modulated according to the user’s sitting or sedentary behaviour. For example, as the user accumulates more“sedentary points”, more product recommendations for the SDPs are modulated from being “Recommended” (or “green”) to “maybe recommended” (or“amber”) and hence the proportion of the graphical user interface element 1 105 occupied by the second bar 1 103B increases. The minimum and maximum lengths of the second bar 1 103B are, respectively, 0% and 100% of the overall length of the GUI element 1 105. In general, the total length of the graphical user interface element remains constant when the length of the second bar 1 103B is updated, i.e. the length of the first bar 1 103A is decreased (or increased) by the same amount as the length of the second bar 1 103B is increased (or decreased).

In some cases, the length of the second bar 1 103B is reset to 0% at a pre-defined time, e.g. at midnight. After this has happened, the length of the second bar 1 103B increases as the user accumulates sedentary points (i.e. his or her CST increases), but cannot be decreased by user undertaking physical activity. However, the user may accumulate points for physical activity (e.g. steps) which can be used to offset sedentary points that are gained later on. Conversely, further sedentary behaviour does not increase the length of the second bar 1 103B when it has reached 100%, but sedentary points may continue to accumulate such that a greater amount of physical activity is required to bring the length of the second bar 1 103B below 100%.

The functionality of the GUI element 1 105 can be extended by allowing the user to define a limit or goal (TGOAL) for the CST, e.g. 4, 6 or 8 hours. The modulation of the product recommendations associated with the SDPs may be de-activated until the user’s CST reaches this limit. Alternatively, the limit may be expressed in terms of a threshold proportion of SDPs, e.g. 40%, being eligible for having their recommendations modulated based on the accumulated CST.

Although the DNA Health Bar 1 105 may normally be displayed on a user’s personal computing device (e.g. a smartphone or tablet), it can also be displayed on the wearable device 100 if it has a suitable display. Alternatively, the wearable device 100 may alert the user to the effects of his or her sedentary behaviour using some other form of visual indication. For example, the visual indicator (e.g. LED) 104 may be illuminated in different colours dependent on whether the proportion of products for which the associated product recommendation is modulated exceeds a certain threshold, e.g. a green colour when the threshold is not exceeded and an amber colour when it is. Alternatively, the visual indicator may be illuminated only when the threshold is exceeded. Illumination of the visual indicator may be triggered by using the one or more inertial sensors to detect that the user has performed a particular motion or motion, such as raising his or her hand.

Figure 12 shows an exemplary relationship or“mapping” between the CST and the proportion of SDPs for which the associated product recommendations are modulated (referred to here as RAMBER), which takes into account the user-defined goal for CST. Starting from a CST of zero, the proportion of SDPs increases linearly with CST until the user-defined goal is reached, and then non-linearly thereafter (although in other examples the proportion of SDPs may increase non-linearly from 0% to 100%). This non-linear dependence may be chosen so that the rate at which 100% is RAMBER reached increases more rapidly with the CST. In the example shown in Figure 4, RAMBER increases exponentially with CST after the user-defined goal has been exceeded. This type of non-linear dependence effectively“penalises” users more heavily when their CST is higher. The mapping also allows the user’s target (TGOAL) for the CST to be converted to a target proportion of product recommendations that are modulated (RAMBER), which can be used to update the GUI element 1 105 to show whether the user has kept below his or her CST goal. For example, in some cases the target proportion is shown explicitly on the GUI element 1 105, e.g. using a vertical line 1 106. Figure 13 shows an example of how the DNA Health Bar 1 105 is updated for a particular user over the course of a 24-hour period. The DNA Health Bar 1 105 may include an indicators for either or both the amount of physical activity (e.g. number of steps) undertaken by the user or the CST. In the figure, the text labels are used as the indicators, although of course forms of graphical indicator could be used.

The DNA Health Bar is initially at 0% (i.e. the length of the second section 303B is zero) and between 7:00-8:00 the user wakes up and takes 5000 steps. During this time, if the user scans a SDP using the wristband 100, then the product recommendation is not modulated, e.g. the indicator of the wristband turns green. In the next hour, the user accumulates 0.5 hours of sitting time (ST) while travelling by train and 1000 steps (by walking from the station, for example). The length of the second bar 303B on the Heath Bar 1 105 is therefore increased to represent 0.5 hours of CST. However, since the amount of CST is less than the 6 hour goal set by the user (as indicated by the vertical line 306 towards the centre of the Health Bar 1 105), none of the SDP product recommendations are modulated yet, i.e. the wristband still shows green if an SDP product is scanned.

Between 9:00 and 13:00, the user is sedentary and accumulates 4 hours of CST, causing the second bar 303B to increase in length. Between 13:00 and 14:00, further steps taken by the user brings the total number of steps accumulated to be greater than 10,000 steps (RSTEP), which causes the CST to decrease, leading to a reduction in the size of the second bar 303B.

Between 14:00 and 18:00, further sedentary behaviour causes the CST to increase above the 6-hour target, TGOAL, set by the user and consequently modulation of the SDP product recommendations is activated, i.e. if the user scans a SDP that is now only“maybe recommended” then the visual indicator of the wristband 100 is illuminated with an amber colour. This indicates to the user that they have not reached managed to moderate their sedentary behaviour sufficiently to stay within the limit they chose.

Between 18:00 and 23:00, the user accumulates further CST, which is not reduced by the further steps that the user has taken, since the user has not accumulated another 10,000 steps. However, if the user were to have taken another 8,000 steps in this period then the CST would have been decreased accordingly. Since the CST is above T GOAL, the length of the second bar 303B increases more dramatically with each additional unit of CST.

Once the user goes to bed, the wristband 100 (for example) may detect that the there is no activity whatsoever from the user and stop increasing the CST (this may also happen if the user removes the wristband 100, for example). Thus, between 23:00 and 0:00, when the user is asleep, the Health Bar 1 105 is not updated. However, once midnight is passed, the statistics for the Health Bar 1 105 are reset so that the user can monitor his or her sedentary behaviour for the next day.

Of course, as is conventional with graphical user interface elements (or“widgets”), the overall size or orientation of the graphic user elements 1 101 , 1 105 may be adjusted to fit the display of the device or the user interface which is used to render the user element 301. The relative positions of the sections 1 102-1 104 or subsections 1 103A, 1 103B may also be adjusted, e.g. the sections may be stacked adjacent to one another, rather than end-to-end, as shown in Figure 3. Colour, shading or another form of visual styling may be used to help visually distinguish the sections 302-304 and/or subsections 1 103A, 1 103B from one another. In some cases, the colour scheme used for the user interface element 1 101 may match the colour scheme used to indicate the product recommendations to the user on the wearable device 100 when a product code is scanned.

In general, the following procedure may be used to obtain product recommendations.

Firstly, a number of health traits associated with potentially poor health are identified, including for example, type-ll diabetes, hypertension, high body mass index (BMI), and high cholesterol. Whether or not a user has any of the these health traits may be determined using genetic testing e.g. based on detecting single nucleotide polymorphisms (SNPs) in a sample of the user’s DNA or RNA, although other forms of diagnostic testing can also be employed, including detection of other types of biomolecule or testing the user’s breath to determine the composition of their microbiome.

Once the“target” health traits have been identified, the effect of diet on each of them is considered. This is done by analysing how the health trait is likely to be affected for each of a number of different categories relating to the nutritional content of consumable products, including any of the following, for example: sugar content, calorific content, carbohydrate content, saturated fat content, total fat content, and salt content. For example, it is known that the risk of type-ll diabetes can be raised by consumption of products that are high in sugar, whilst salt-rich diets are linked to hypertension. The relationship between each category and each health trait can be quantified by assigning a trait-dependent coefficient to each of the categories. Taking hypertension as an example, large coefficients may be assigned to salt content and fat content, whilst small (or even zero) coefficients may be assigned to calorific content and carbohydrate content.

For each identified health trait, the trait-dependent coefficients can be used to calculate a score that indicates the expected effect of a product based on its nutritional content, i.e. how many grams of salt, saturated fat, etc. the product contains. For example, to calculate a score that is indicative of the adverse effect of a product on hypertension, the number of grams of salt in the product can be multiplied by the coefficient for salt content. Mathematically speaking, the score for each trait may be determined by taking a scalar product of a vector of the trait-dependent coefficients with a vector of nutritional information for each product. Of course, more sophisticated ways of calculating the scores can also be used, such as using a dose-response curve determined for a health trait to model the likely effect of the different types of nutritional category on that trait. The score may also take into account other factors, such as the type of product (e.g. confectionary, biscuits, breakfast cereals etc.) and the typical portion size for the products. These other factors may be used to adjust the scores for differences in how different products are consumed, e.g. if a product is identified as typically being consumed as a“treat” then its score may be reduced so that its recommendation may be more positive than it would otherwise have been.

After the scores have been calculated for each product, the products are ranked in descending order according to their score for a particular health trait. A subset of the products is then chosen by selecting products which have a ranking which is higher than a threshold ranking, e.g. products which are in the top 50% or 30% or 10% of products based on their score of the health trait. These products may be assigned a “negative” recommendation, such as“not recommended”, so that when the wristband 100 is used to read the barcode of the product, the indicator (LED) 106 turns red (for example). The remaining products are assigned to a different category, e.g. “recommended” (which would be indicated to the user with a green colour). Thus, the products are allocated different recommendations according to their ranking for a particular trait. The value of the threshold ranking or“cut off” which is used for allocating the recommendations varies from user to user and is determined based on how susceptible the user is to the particular trait (as determined by genetic testing, for example).

The product recommendations for each health trait are combined to determine an overall product recommendation for each product. This can be done in a number of ways, such as determining that the product is“not recommended” if the product is“not recommended” for any of the traits or, alternatively, if the product is “not recommended” for more than one of the traits. The flag indicating whether the product recommendations may be modulated can also be set or unset for each product based on the same or similar kinds of rule. For example, when the product recommendations are generated, a“tag” may be associated with each of the product codes. One value of the tag (e.g. 0) may be used to denote non-SDPs whilst other values may be used to denote the product codes associated with SDPs. In some cases, the CST can be used to define an“activation threshold” value for the tag such that all products with a tag value that is higher than the threshold value are modulated.

The subset of products for which the associated recommendation may be modulated according to the user’s sitting behaviour can be chosen in a number of ways. In general, the modulation is not applied to all types of food product because to do so may cause“healthy” products such as vegetables to be“maybe recommended” based on the user’s sitting behaviour. The subset may be chosen based on nutritional data associated with each product. In one example, the products that are classified as being“recommended” are ranked according to their calorific content and the top 50%, 30% or 10% of the products with the highest calorific content are selected for potential modulation. In some embodiments, the cut-off percentage for selecting the products this way can be adjusted by the user in order to increase or decrease the number of products for which the modulation is applied. The number of products (in the subset mentioned above) for which the associated product recommendations are modulated is determined from the CST as discussed above in relation to Figure 4.

In general, the threshold rankings mentioned above, and the rules for determining the overall product recommendations, strike a balance between allowing users to choose products they wish to consume and dissuading them from consuming at least some of the products that are most likely to have a negative impact on their health. This balance is advantageous because the user is“nudged” into making choices that are better for him or her in the long term. Modulating the product recommendations based on the user’s sitting and/or other sedentary behaviour provides an additional“nudge” so that, for example, a user may notice that a product which is normally “recommended” is only “maybe recommended” (or recommended as being “discretionary”), if the user has been sitting for long periods during a particular day. Using a modulation which does not make the products“not recommended” may be particularly advantageous because it does not penalise the user too harshly for circumstances which may be out of their control, e.g. if the user has been sitting for a long time because they have had to travel a large distance.

The steps in a method of alerting a user to the negative effects of sedentary behaviour are as follows:

Step 1 : Store in a memory of a computer device, product codes for consumable products and data indicating respective product recommendations or from which respective product recommendations can be derived.

Step 2: Obtain motion data from one or more inertial sensors worn by the user.

Step 3: Monitor the behaviour of the user by using the motion data to identify periods when the user is in a sitting position or other sedentary state.

Step 4: Based on the behaviour of the user, selecting from at least a subset of the product codes, product codes for which the respective product recommendation is to be modulated Step 5: Control a visual indicator to provide a visual indication dependent on the number of selected product codes.

Figure 14 shows an exemplary graphical user interface 701 incorporating the DNA Bar 1 101. Figure 15 shows an exemplary graphical user interface 801 incorporating the Health Bar 1 105. Each of these GUIs may be rendered

An exemplary system comprises a personal computing device (such as a smartphone, smartwatch, tablet computer, or desktop device) and the wearable device 100. The personal computing device comprises a transceiver for exchanging data with the wearable device 100 (e.g. via a wired or wireless connection), a memory for storing data, a processor for processing data received from the wearable device 100, and for controlling a display to display the DNA Bar 301 and/or the Health Bar 1 105. In use, the wearable device 100 may transmit motion data to the personal computing device, which monitors the behaviour of the user by using the motion data to identify periods when the user is in a sitting position or other sedentary state and then, based on the behaviour of the user, selects product codes for which the respective product recommendation is to be modulated. Alternatively, the monitoring and selecting may be performed by the wearable device 100 itself, in which case the wearable device 100 transmits data indicative of the selected product codes to the personal computer device. The processor updates the display to provide a visual indication dependent on the number of selected products, e.g. by updating the Health Bar 1 105.

The memory may the store the product codes and data indicating respective product recommendations or from which respective product recommendations can be derived. The product codes and data may be downloaded onto the wearable device 100, e.g. after the product codes and/or product recommendations have been updated.

Although the above description has focussed on sitting or other sedentary behaviour, the above aspects may also (or alternatively) equally be applied to other facets of a user’s lifestyle and/or to the user’s environment. For example, as described above the wearable device may, in some instances, comprise a UV light sensor, which can be used to monitor the user’s exposure to UV light (of course body-worn sensors which are not part of the wearable device 100 can also be used). In this case, the user’s product recommendations may be adjusted according to his or her behaviour in terms of UV light exposure, e.g. time spent in direct sunlight. The wearable device 100 or personal computing device may then be configured to provide a visual indication (e.g. using a DNA Health Bar 305) of the number of product recommendations that are to be modulated based on periods of UV light exposure. This allows the user to be alerted to the potentially harmful effects of UV light exposure (e.g. by recommending products which lead to higher levels of vitamin E). Similarly, the system can be configured to provide the user with feedback (in terms of the number of modulated product recommendations) on potentially harmful effects of the composition of their environment based on data obtained using body-worn sensors for air quality or pollutants (e.g. NOx or particulates).

In another example, product recommendations may be modulated according to the duration and/or intensity of a user performing a particular physical activity (or physiological function), such as running (based on data obtained using one or more sensors, e.g. a pedometer). For example, some users may be identified by genotyping as being likely to have low bone density and it may therefore be appropriate to modulate some of the product recommendations so that certain types of footwear are favoured in order to mitigate the effects of high impact activities on the user’s skeletal system. Providing the user with a visual indication of this modulation may alert them to the need for remedial action, such as taking up a lower impact activity.

In some implementations, multiple DNA Health Bars 1 105 (or other forms of visual indicator) may be provided (e.g. in the form of a user interface“panel”) to allow a user to track the relative contribution of various different user lifestyle and user environment factors to the number of product recommendations that are to be modulated. For example, a user who spends less time sitting by walking may only see a modest decrease (or even an increase) in the number of modulated product recommendations if he or she walks in direct sunlight or in a polluted area. Alerting the user to this issue using multiple visual indicators may therefore allow the user to adjust his or her behaviour more appropriately.

It will be appreciated by the person of skill in the art that various modifications may be made to the above-described embodiments without departing from the scope of the present invention. For example, whilst the principle embodiments described have been in the form of a wearable band, the system may be implemented in any suitable format, for example as a holdable cartridge, keyring, pendant, or a smartphone, or any combination of such formats. It is further noted that the data stored in the system may be derived from biological information obtained from an analysis of a biological sample provided by the user and samples provided by other users. This set of users could be the members of a family. The data store then contains a pool of common data that can be used to provide the best recommendations for all family members.