DESCRIPTION

TECHNICAL FIELD

The present invention relates to the field of electronic systems. In greater detail, the present invention relates to a method and system of assisting the execution of physical exercises.

BACKGROUND

The use of devices to monitor a user's physiological parameters is widespread. In particular, these devices are wearable electronic devices configured to detect various physiological information of the user such as heartbeat, movement in space, waking and sleeping periods, etc.

This information has enabled the development of methods and systems capable of monitoring and evaluating the user's physical activity.

For example, US 10,293,207 discloses a system capable of identifying predetermined “events” based on information provided by a wearable device. These events can be traced back to a movement of interest such as the execution of an intense physical activity. The system comprises evaluating the user's performance based on the analysis of the energy released by the user during the identified event.

In other words, methods and systems are known that allow monitoring and classifying the physical activities executed by a user. In addition, these methods and systems are configured to provide additional information such as the energy generated during the execution of a recognized physical exercise or the calorie consumption associated with the execution of the physical exercise.

However, these known solutions do not allow obtaining any useful information on the quality of execution of an exercise. This is an important aspect related to physical activity. In fact, the benefits deriving from the physical activity are conditioned by the fact that it is executed correctly, respecting the abilities and physiology of the user. On the other hand, executing physical exercises incorrectly intuitively leads to conditions of muscle fatigue and even serious injuries, thus achieving the opposite of the desired result of improving one's physical/motor skills.

The Applicant has identified some known methods and systems that attempt to solve this problem.

For example, US 10,789,708 proposes a method and a system configured to identify incorrect movements of a user's torso while running.

US 10,729,964 describes, instead, a system in which a user is monitored by means of one or more electronic devices, including a wearable electronic device. The system is configured to identify the occurrence of events that correspond to one or more previously selected patterns. The system is also configured to provide a performance evaluation based on a degree of match between the identified activity and the model stored by the system.

WO 2016/157217 describes a system for tracking a user's activity comprising a wearable device provided with a movement sensor and a force sensor. The start and the end of a physical activity are determined through the analysis of the data provided by the wearable device. In addition, data analysis is used to identify the exercise executed and to predict the execution accuracy of the exercise.

US 2017/337033 describes a method for selecting music based on the detection of a physical activity performed by a user, e.g. running, swimming, hiking, etc. A wearable device uses data provided by one or more biometric sensors to determine that the user has started a particular physical activity and, if so, to play music, for example through a music playback device or software.

US 2021/008413 describes a system for analysing the execution of a physical exercise by a user and providing feedback and recommendations regarding the execution of the exercise. Through a stream of data provided by one or more sensors associated with the user, in particular comprising video cameras, the user's movement is analysed over a period of time in order to detect the execution of one or more poses. On the basis of the poses detected, the exercise executed is identified and an indicative score of the correctness of execution of the exercise is assigned.

The Applicant has noted that the known solutions have limited success in evaluating the quality of execution of a physical exercise. In particular, the known solutions require multiple types of sensors or, more generally, multiple monitoring devices, both wearable and non-wearable, in order to correctly evaluate the quality of an exercise, which makes the implementation of such systems and methods complex and costly. In addition, the Applicant noted that the known solutions are capable of evaluating an extremely limited number of exercises typically selected a priori.

OBJECTS AND SUMMARY OF THE INVENTION

An object of the present invention is to overcome the drawbacks of the prior art.

In particular, it is an object of the present invention to provide a simple and efficient method and relative system capable of evaluating the quality of execution of a varied type of physical exercises in a reliable manner.

Furthermore, it is an object of the present invention to provide a method and a relative system configured to promptly alert a user when an exercise is executed incorrectly and/or sub-optimally.

Finally, it is an object of the present invention to provide a method and a relative system configured to signal to a user when to stop repeating an exercise in order to prevent injuries.

These and other objects of the present invention are achieved by a system incorporating the features of the annexed claims, which form an integral part of the present description.

According to a first aspect, the present invention is directed to a method of assisting the execution of physical exercises. The method comprises the steps of: by means of a wearable electronic device provided with a plurality of inertial sensors worn by a user, acquiring measures of a set of three movement components transverse to each other, each parallel to an axis of a Euclidean reference system; determining whether there is a main movement component of said set of three, i.e. a movement component substantially greater than the other movement components; if so, comparing the set of three movement components with a set of set of threes of reference movement components to identify a physical exercise executed when a match between the set of three measured movement components and a set of three reference movement components of the set of reference movement components is identified - advantageously, the set of set of threes of reference movement components comprises a plurality of set of threes each measured during the execution of a corresponding known physical exercise; calculating a quality index of the execution of the physical exercise as a function of a deviation of a movement dynamic resulting from the set of three movement components from a direction parallel to an axis or lying in a reference plane of the Euclidean reference system parallel to the main movement component; verifying whether the quality index of the execution of the physical exercise is below a threshold value, and if so, generating an error signal, wherein the error signal comprises a message understandable to a user wearing the wearable electronic device.

Thanks to this solution, it is possible to obtain an objective evaluation of the quality of execution of any exercise executed by the user in a completely automatic way and with extremely limited hardware resources. In particular, only inertial sensors, e.g. linear accelerometers, are basically required to execute the method correctly.

Furthermore, it is possible to identify a first category of physical exercises that can be reliably evaluated. In particular, the Applicant has observed that the quality of physical exercises having a main movement component can be assessed in a simple manner by evaluating a level of movement deviation or match with respect to an ideal direction to which the movement component should be aligned.

Finally, evaluating the execution of the physical exercise as a function of the deviation of the movement dynamic from the axis or plane of reference associated with the main movement component makes it possible to evaluate the execution of a physical exercise independently of time. This gives a wide versatility to the evaluation method that can be applied both on a single execution of the physical exercise and on a series of repetitions of the physical exercise.

In one embodiment, the step of determining whether a main movement component exists comprises calculating the ratio between each movement component and the sum of the movement components, and identifying the main movement component as the movement component for which said ratio equals or exceeds a predetermined minimum value.

In one embodiment, the movement components comprise one between speed, rotation speed, acceleration, and movement path.

Preferably, the movement components correspond to a set of three speed v/, v and v/ having directions parallel to a set of three reference axes XA, YA and ZA of the Euclidean reference system originating in the fitness tracker.

The Applicant has found that the use of speed is easier to use than the acceleration value, as it is characterised by a simpler dynamic. Otherwise, the use of the position as a function of time would have introduced an excessive error unless the processing complexity and, therefore, the hardware resources required to perform the method were substantially increased, in particular by having to perform a double integration operation on the acceleration measures provided by the inertial sensors of the wearable device.

Even more preferably, the quality index of the execution of the physical exercise, or directional performance index, is calculated as:

IPD = 100 wherein IPD is the directional performance index, D(v/), D(v ) and D(v/) are the statistical dispersion indices, preferably Poisson dispersion indices, of the speeds v/, v and v , and D(i¾? _{y z} ) is the statistical dispersion index, preferably the Poisson dispersion index, of the speed corresponding to the main movement component between the speeds.

Additionally or alternatively, a directional error percentage, defined as err = 1 — IPD, can be calculated. However, the directional performance index can be calculated starting from rotation speed, acceleration, or trajectory of movement mutatis mutandis.

In other words, the directional performance index, in generic terms, can be defined as:

_ ^D (jx,y,z) _

IPD 100

D(fx )+D(fy )+D(vf _z ^A Y wherein f/, f and f/, indicates a type of movement component selected from linear speed, rotation speed, acceleration, or movement path, i.e., an indication of position as a function of time.

In one embodiment, in the case of a main movement component aligned to a plane, i.e. the main movement component comprises a pair of movement components, the directional performance index is defined in percentage form as:

IPD 100 wherein D(f _x ^A _yz ) is the statistical dispersion index, preferably the Poisson dispersion index, of the movement component that does not belong to the main movement component.

In the case where the main movement component comprises two speeds parallel to the axes defining the plane to which the main movement component is aligned, the directional performance index is calculated as:

IPD 100

The Applicant determined that defining the directional performance index in this way allows for a quick and efficient calculation thereof. Furthermore, the directional performance index calculated in this way is particularly reliable and independent of the peculiarities or experience of the athlete executing the physical exercise. Furthermore, the calculation of the directional performance index does not require the use of measures, speeds or other reference information.

In addition, when the physical exercise executed by the user does not have a main movement component, it the method comprises:

- identifying a type of physical exercise executed by the user on the basis of the acquired measures referring to the execution of at least one physical exercise executed, and

- calculating a further quality index of execution of the physical exercise, or correlation performance index, indicative of a comparison between the correlation between pairs of acquired measures referring to the execution of the at least one physical exercise and the correlation between pairs of a set of reference measures.

Thanks to this solution it is possible to evaluate the quality of execution of exercises that are not characterized by a main movement component aligned along a specific direction in a simple and reliable way.

Preferably, the step of calculating the quality index of execution of the physical exercise, a correlation performance index in this case comprises: calculating a set of three correlation coefficients between pairs of speed as: wherein corr(v _x , V _y ), corr(v _x , v _z ) and corr(v _y ,v ^A ) denote the correlation operation between the pairs of speed v _x , v ^A , v _x , v _z and v ^A , v _z , respectively, and calculating the correlation performance index as: wherein IPC is the correlation performance index, and f _xy , f _xz and f _yz are reference correlation coefficients associated with the identified type of physical exercise.

Even more preferably, the correlation between pairs of speed is calculated as a Pearson linear correlation index.

For example, the reference correlation coefficients are determined by calculating a set of three correlation coefficients between measured pairs of speed during the execution of at least one execution of the exercise by a professional athlete.

Again, the directional performance index can be calculated from rotation speed, acceleration, or trajectory of movement mutatis mutandis.

In other words, the correlation performance index, in generic terms, can be calculated starting from correlations among movement components other than speed, for example movement components selected from rotation speed, acceleration, or movement trajectory, i.e.: wherein corr(f ^A ,f ^A ), corr (f ^A ,f _z ^A ) and corr[f ^A ,f ^A ) denote the correlation operation between the pairs of movement components considered.

The Applicant has determined that the movement components, such as the speed, of a physical exercise are correlated to each other as part of the same physical exercise. In other words, the analysis of the correlations among the movement components makes it possible to obtain reliable information on the dynamic of the movements associated with the execution of the physical exercise in a static manner. Consequently, it is possible to evaluate the quality of execution of a physical exercise extremely easily and quickly by comparing the correlations between calculated movement components and a set of reference correlations associated with the physical exercise executed.

In one embodiment, further comprising the step of determining the threshold value used to verify the quality index of execution of the physical exercise on the basis of a level of competence of the user wearing the wearable electronic device.

In addition or alternatively the threshold value may be based on a load borne by the user wearing the wearable electronic device during the execution of the exercise.

Thanks to this solution it is possible to adapt the threshold value and therefore the identification of the execution of an incorrect exercise in a personalised manner, based on the competences and/or the amount of load borne by the user in order to provide more fitting indications to the actual condition of execution of the exercise.

Additionally or alternatively, it is provided to control a trend of the value of the directional performance index or of the correlation performance index during a plurality of consecutive executions of the physical exercise. In the event that said trend is lower or deviates from an expected trend range of values, to generate an alert message comprising an intelligible indication to the user to stop the activity of execution of the physical exercise. Thanks to this solution, it is possible to identify fatigue conditions in the user at an early stage that lead to a progressive reduction in the quality of execution of the physical exercise and thus increase the risk of injuries.

In one embodiment, the measures provided by the sensor assembly refer to a Euclidean reference system defined by a first set of three axes originating in the wearable electronic device, and wherein the method comprises converting the measures so that they are referred to a Euclidean reference system defined by a second set of three axes originating in the wearable electronic device and oriented so that an axis is parallel to the direction of the acceleration force of gravity.

In this way, the measures provided by inertial sensors can be analysed with greater precision and simplicity.

A different aspect of the present invention relates to a system of assisting the execution of physical exercises comprising: a wearable electronic device and a processing unit. The wearable electronic device comprises a plurality of inertial sensors and is configured to transmit a plurality of measures made by the inertial sensors to the processing unit. The system is configured to execute the method according to any of the above embodiments.

A different aspect of the present invention relates to a system of assisting the execution of physical exercises comprising a wearable electronic device. The wearable electronic device comprises a plurality of inertial sensors and a control module and is configured to execute the method according to any one of the embodiments described above.

Further features and advantages of the present invention will be more evident from the description of the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention is described hereinbelow with reference to certain examples provided by way of non-limiting example and illustrated in the accompanying drawings. These drawings illustrate different aspects and embodiments of the present invention and reference numerals illustrating structures, components, materials and/or similar elements in different drawings are indicated by similar reference numerals, where appropriate.

Figure 1 schematically illustrates a system of assisting the execution of physical exercises according to an embodiment of the present invention;

Figure 2 is a block diagram of a wearable electronic device included in the system of the present invention;

Figure 3 is a block diagram of a remote management unit included in the system of the present invention, and

Figure 4 is a schematic representation of a user wearing a wearable electronic device while executing an exercise;

Figure 5 is a flow chart of a method of assisting the execution of a physical exercise according to an embodiment of the present invention;

Figures 6a and 6b are graphs of the speeds along three reference axes as a function of time during the execution of a physical exercise correctly and incorrectly, respectively, and

Figures 7a and 7b are graphs of the correlations between pairs of speed along three reference axes as a function of the measurement samples acquired during the execution of a physical exercise correctly and incorrectly, respectively. DETAILED DESCRIPTION OF THE INVENTION

While the invention is susceptible to various modifications and alternative constructions, certain preferred embodiments are shown in the drawings and are described hereinbelow in detail. It must in any case be understood that there is no intention to limit the invention to the specific embodiment illustrated, but, on the contrary, the invention intends covering all the modifications, alternative and equivalent constructions that fall within the scope of the invention as defined in the claims.

The use of lor example”, “etc.”, “or” indicates non-exclusive alternatives without limitation, unless otherwise indicated. The use of “includes” means “includes, but not limited to” unless otherwise indicated.

With reference to Fig.1 a system of assisting the execution of physical exercises according to an embodiment of the present invention, in short the system 1 in the following, comprises a wearable electronic device, a fitness tracker 2 in the example considered, and a remote processing unit, for example a server 3.

The fitness tracker 2 comprises a band 10, to allow a user to wear the fitness tracker on the wrist, and a casing 20, preferably waterproof, which encloses an electronic circuit 30 (illustrated in Figure 2). For example, the casing 20 is formed by a pair of shells constrained to each other by means of mechanical mutual fasteners and is provided with an input and output user interface 21 , for example a screen comprising a touch sensor.

The electronic circuit 30 of the fitness tracker 2 comprises a sensor assembly 31 , a control module 32 - preferably, comprising a microcontroller or a microprocessor -, a memory module 33, a communication module 34 - capable of exchanging data with the server via a wireless communication channel W - and a power supply module 35 - comprising a battery. In particular, the control module 33 is connected to at least the modules 31 , 32, 34 and is configured to govern the operation thereof. In addition, the power supply module 35 is connected to the modules 31 , 32, 33 and 34 and is configured to deliver operating electrical power thereto.

In embodiments of the present invention, the sensor assembly 31 comprises a plurality of inertial sensors S1 -Sn, for example a set of three acceleration sensors. In detail, each acceleration sensor is configured to provide acceleration measures along a respective direction. Preferably, the three directions are orthogonal to each other so as to define a three-dimensional Cartesian space x, y, z centred on the fitness tracker 2. In one embodiment of the present invention, the acceleration sensors each comprise an accelerometer. Preferably, the sensor assembly 31 also comprises a set of three angular speed sensors, wherein each angular speed sensor is configured to provide rotation speed measures about a respective rotation axis. Preferably, each rotation axis corresponds to one of the directions x, y, z with respect to which the linear accelerations are calculated. In one embodiment, the angular speed sensors comprise a gyroscope. For example, the fitness tracker 2 comprises an inertial platform or Inertial Measurement Unit (I MU).

With reference to the example in Figure 4, the measures provided by the sensor assembly 31 refer to a Euclidean reference system defined by the axes XT, y-r, and ZT originating in the fitness tracker 2 worn by a user U. In other words, the measures provided by the sensor assembly refer to a Euclidean space, specifically a three-dimensional space, where each point of the space is defined by three coordinates referring to as many axes orthogonal between them. In the considered embodiment, these measures are converted into an ^absolute” Euclidean reference system defined by the axes XA, y _A and ZA are determined so that the elevation axis ZA is aligned with the direction of the acceleration force of gravity g - which is determined through the analysis of the measures provided by the set of three accelerometers in a per se known manner and therefore not described in detail. Finally, the communication module 34 is configured to establish the communication channel W with the fitness tracker 2 and the server 3 through a communication network (not illustrated). For example, the communication module 34 comprises a wireless modem configured to connect to an intermediate device such as a WiFi modem, a Bluetooth® modem, and/or a base station of a cellular network (not illustrated) to establish a connection to the communication network through which the data are transferred to the server 3.

The server 3 is configured to exchange data with the fitness tracker 2, in particular to store and process the data received from the fitness tracker 2 and transmit a result of such processing to the fitness tracker 2.

For this purpose, the server 3 comprises a processing module 41 - formed by one or more of microcontrollers, microprocessors, general purpose processors (CPUs) and/or graphic processors (GPUs), DSPs, FPGAs, ASICs, edge computing resources, etc. -, a memory module 42 preferably configured to store large amounts of data, for example, organised in one or more databases and a communication module 43 configured to establish the communication channel W with the fitness tracker 2 through a communication network (not illustrated). For example, the communication module 33 comprises a modem, preferably connected to a telecommunication network. Finally, the server 3 comprises one or more ancillary modules (not illustrated), such as one or more power supply modules configured to supply the necessary operating energy to the processing module 41 , to the memory module 42 and to the communication module 43.

The server 3 is configured to implement a software application 4 configured to evaluate a quality of execution of the physical exercises performed by a user wearing the fitness tracker 2 and provide an alert message when an incorrect execution of the physical exercises is identified and, preferably, indications on how to correct the execution of the physical exercises.

In the preferred embodiment, the software application 4 implements a procedure 100 for assisting in executing physical exercises of which Figure 4 is a flowchart.

The procedure 100 initially comprises acquiring a plurality of measures acquired by the sensor assembly 31 of the fitness tracker 2 (block 101). Preferably, the measures are acquired continuously and transmitted to the server 3 substantially in real time, although nothing prohibits transmitting groups of acquired measures at consecutive measurement time ranges. Preferably, the fitness tracker 2 is configured to sample the acceleration values measured by the sensor assembly 31 with a predetermined sampling frequency and transmit a stream of samples or groups of samples of measures acquired at consecutive time instants.

The acquired measures are first converted from acceleration measures into speed measures (decision block 103) and then analysed to determine whether the user is executing a physical exercise (decision block 105). In the considered embodiment, the measures are analysed by means of a classification algorithm, for example based on artificial intelligence (Al), configured to identify the execution of a physical exercise, distinguishing it from a state of ‘test” - or, more generally, from the non-execution of a physical exercise. For example, a movable observation time window Mo is defined and the measures obtained from the sensor assembly 31 are processed to extract distinctive features associated with the execution of a physical exercise - for example, statistical features, in the frequency domain, of autocorrelation and/or of mixed correlation. The distinguishing features are then provided as input to a machine learning (ML) algorithm which is trained to determine whether the user is executing a physical exercise. Preferably, the ML algorithm is of the XG Boost type.

Even more preferably, the ML algorithm is configured to identify a start instant and an end instant of each physical exercise executed. In this way, it is possible to identify each repeated execution, or “repetition” for short, of a specific physical exercise as is common during a training session involving free-body exercises, with weights/ballasts and/or other gymnastic equipment (such as parallel bars, rings, bars, etc.).

In one embodiment, the observation time window Atq has a duration greater than or equal to the duration of the physical exercise considered. Furthermore, a translation approach of the observation time window Atq is applied with overlap, i.e. a part of the measures provided by the sensor assembly 31 or speed v _x ^A , v and Vz ^A included in the observation time window Atq considered at a generic time f, correspond to the measures provided by the sensor assembly 31 or speed v _x ^A , v _y ^A and v _z ^A included in the observation time window Atq considered at a previous generic time instant t-1. For example, the overlap between two consecutive windows is comprised between 60% and 80%, preferably equal to 70%, of their time span.

If the execution of a physical exercise is not identified (output branch N of block 105), the analysis of new measures acquired by the sensor assembly 31 of the fitness tracker 2 continue - in other words, the procedure returns to block 101 .

On the other hand, if the execution of a physical exercise is identified (output branch Y of block 105), a number of consecutive repetitions of the execution of a physical exercise is counted (block 107) and, in series or parallel, the executed physical exercise is classified (decision block 109).

In other words, the counting envisages identifying a sequence S comprising one or more executions of the same physical exercise, for example, a so-called “repetition set”, that is a predetermined number of consecutive executions of the same physical exercise and an index of the quality of the execution of the physical exercise executed is provided. In contrast, the exercise classification is performed in an automated manner by an ML algorithm suitably trained to identify the type of exercise executed by the user, e.g. on the basis of distinctive features extracted from the measures provided by the fitness tracker 2.

In the example considered, the physical exercise classification includes determining whether a dynamic of the physical exercise executed by the user includes a main movement component parallel to a specific direction or lying in a specific plane. In detail, a movement component of the dynamic of the physical exercise is defined as a displacement, a speed and/or an acceleration that are parallel to a respective reference axis XA, yA and ZA of the fitness tracker 2 during the execution of the physical exercise. Alternatively, movement component means a displacement lying in a specific plane - in the example considered, one of the three planes a, b or y, each defined by a pair of axes XA, y _A and ZA as appreciable in Figure 4. Accordingly, the main movement component comprises the two movement components parallel to the axes defining the plane in which the main movement component lies. The main movement component is the movement component that is substantially greater than the others, for example, in terms of absolute maximum value or more preferably of absolute mean value.

In the considered embodiment, the main movement component is identified on the basis of a comparison of the three movement components considered. Preferably, the comparison is performed by determining an important ratio between a movement component and the sum of the three movement components. In this case, the main movement component is identified as the movement component for which the above- mentioned importance ratio has a value equal to, or greater than 0.75, preferably equal to, or greater than 0.85.

Similarly, the main movement component is considered parallel to a plane, i.e. it comprises a pair of movement components parallel to two of the axes XA, y _A and ZA, when by applying the comparison it results that the sum of the importance ratios referred to two movement components is equal to or greater than 0.85.

Examples of physical exercises that have a dynamic with a main movement component along the axis ZA comprise, but are not limited to, shoulder press, squat, snatch and their variants for which the main component (displacement, speed or acceleration) is oriented along an elevation direction. Again, physical exercises such as walking lunge have a main component oriented along the axis yA - or the plane g - while lateral lunges have a main component oriented along the axis yA - or the plane b.

In one embodiment, the movement components, i.e., the measures or a reprocessing of the acquired measures, are compared to a set of movement components, i.e., model, reference measures stored in a lookup table or a database. For example, the speed measures v _x ^A , v _y ^A and v _z ^A , aligned with the axes XA, y _A and ZA, detected during a single execution of a physical exercise, a portion of the sequence S, the entire sequence S, or a mean of the values of the executions of the physical exercise of the sequence S are compared with set of threes of reference speed values included in the lookup table, where each set of three reference speed values is associated with physical exercises having a respective main movement component (thus one set of three reference speed values for each axis XA, y _A and ZA) or no main movement component. When the speed measures v _x ^A , v _y ^A and v _z ^A correspond to, or are included within a tolerance range from, a set of three reference values, a physical exercise with main movement component or without main movement component associated with the set of three reference values is identified.

When a type of physical exercise is identified that comprises a main movement component, i.e. a main speed v _x ^A , vy* and v _z ^A , along a specific direction or lying in a specific plane a, b or g (output branch Y of block 109), a directional performance index IPD (block 111) is calculated. In detail, the directional performance index IPD is a measure of the quality of execution of the physical exercise. In other words, the directional performance index IPD indicates the extent to which the physical exercise is executed while maintaining a correct form of execution characterised by a main movement component substantially concentrated along a predetermined direction.

In the preferred embodiment, the directional performance index IPD is a ratio between a statistical dispersion index, preferably the Poisson dispersion index, (or relative variance, i.e. a ratio between the variance and the arithmetic mean of a considered variable) of the main movement component and the sum of the statistical dispersion indices of all the movement components calculated for the sequence S of executions of the physical exercise. For example, considering a physical exercise in which the main speed is the speed parallel to the axis ZA, the directional performance index IPD is defined in percentage form as:

IPD = 100 wherein D(v/), D(v ) and D(v/)are the statistical dispersion indices, preferably Poisson dispersion indices, of the speeds v _x ^A , v _y ^A and v _z ^A , respectively, calculated for one or more exercises included in the sequence S. In particular, the closer the directional performance index IPD is to 100% the better the quality of execution of the physical exercise.

Advantageously, in the case of a main movement component aligned to a plane - for example the plane g in the case of lunge -, i.e. the main movement component is a pair of movement components - for example two speeds parallel to the axes y and z in the case of lunge -, the directional performance index IPD is defined in percentage form as:

IPD = 100 i.e. on the basis of the incidence of deviation of the movement component not belonging to the plane parallel to the main movement component. In other words, it is evaluated whether a movement dynamic - that is, the movement resulting from the set of movement components - of the physical exercise executed by the user comprises, for a correct execution, a main movement component parallel to a specific direction of one of the axes XA, y _A and ZA or lying in a specific plane - in the example considered, one of the three planes a, b or y, each defined by a pair of axes XA, y _A and ZA as appreciable in Figure 4.

As will be apparent to the person skilled in the art, the directional performance index IPD is independent of time and can be calculated with respect to a single execution of the exercise, the entire set of repetitions included in the sequence S and/or a subset of repetitions included in the sequence S - for example, one or more initial repetitions and one or more final repetitions of the sequence S. The choice of the number of executions of exercises on which the directional performance index IPD is evaluated depends for example on the desired response speed of the system and/or on the computational load desired or manageable by the server 3.

The directional performance index IPD is then compared with a correlation performance threshold value (decision block 113), for example comprised between 90% and 98%, preferably equal to 96%. In other words, it is verified whether the movement components highlight an (excessive) deviation with respect to the direction or to the main plane, that is if the main movement component is excessively reduced.

In the event that the directional performance index IPD is equal to or greater than the correlation performance threshold value (output branch Y of block 113), the physical exercise is considered to have been executed correctly, as in the case of Figure 6a illustrating a graph of the speeds v _x ^A , v _y ^A and v _z ^A as a function of time during the execution of a shoulder press set, wherein the speed v _z ^A along the main direction ZA of execution of the physical exercise is substantially greater than the speeds v _x ^A and v _y ^A developing in the plane a. Optionally, the method comprises transmitting a message of correct physical exercise to the fitness tracker 2 containing information confirming the correct execution of the physical exercise to the user U, for example comprising the value of the directional performance index IPD (block 115). In other words, the user U receives an evaluation of the manner of executing the physical exercise after each set of repetitions of the physical exercise.

Conversely, when the directional performance index IPD is lower than the correlation performance threshold value (output branch N of block 113), the physical exercise is considered to have been executed incorrectly, as in the case of Figure 6b illustrating a graph of speeds v _x ^A , v and v _z ^A as a function of time during the execution of a shoulder press set, wherein the speeds v _x ^A and v developing in the plane a lead to an excessive reduction of the directional performance index IPD - for example, the IPD of the case in Figure 6b is approximately equal to 73% - or, dually, a percentage of directional error (defined as err = 1 - IPD) exceeds a threshold value. Accordingly, a message of incorrect physical exercise is transmitted to the fitness tracker 2 containing information that warns the user U of an incorrect execution of the physical exercise, for example comprising the value of the directional performance index IPD (block 117).

Returning to decision block 109, when a type of physical exercise is identified that does not comprise a main movement component, i.e. a speed v _x ^A , v and v _z ^A , along a specific direction or lying in a specific plane (output branch N of block 109), the exact type of exercise executed is identified (block 119) and a correlation performance index IPC is then calculated (block 121 ).

For example, measures are analysed by means of a classifier algorithm, for example, based on an appropriately trained Al, to identify the type of physical exercise executed by the user. Preferably, the Al comprises the ML algorithm described above which also classifies the exercises comprising a main movement component. In other words, the ML algorithm is configured to classify both physical exercises having a main movement component and the physical exercises not comprising a main movement component.

In general, the identification of the type of physical exercise performed by means of Al is performed in a manner substantially corresponding to what is described above in relation to the identification of the execution of a physical exercise mutatis mutandis. For example, an artificial percetron multilayer neural network or MLP (acronym for MultiLayer perceptron) trained to accurately identify the exact type of physical exercise executed by the user is used.

Furthermore, also in this case, the correlation performance index IPC is a measure of the quality of the technique of execution of the physical exercise evaluated on a sequence S of two or more executions of the same physical exercise, preferably, a so-called “repetition set” or a predetermined number of executions of the same physical exercise substantially in continuity. Unlike the directionality performance index IPD, the correlation performance index IPC is based on a relationship among the correlation values v _x ^A , v _y ^A and v _z ^A . In the preferred embodiment, correlation coefficients among pairs of speed are calculated: wherein corr( A, B) indicates the correlation operation between a variable A and a variable B. In the embodiment considered, the correlation operation comprises calculating a correlation index, or coefficient, preferably the Pearson correlation index, i.e.: wherein / is an index ranging from 1 to N with N being an integer and corresponding to the number of measures (i.e. samples) considered. Consequently, the correlation coefficients r _xy , r _xz and r _yz are values comprised between -1 and +1 .

Finally, the correlation performance index IPC is calculated as a combination of the correlation coefficients compared with reference correlation coefficients obtained by analysing the measures of a plurality of correctly executed physical exercises. In the embodiment considered, the correlation performance index IPC is calculated as a percentage as follows: wherein r _xy , r _xz and r _yz are reference correlation coefficients associated with the type of physical exercise identified (at block 119). Also in this case, the closer the value of the correlation performance index IPC is to 100% the better the technique of execution of the physical exercise will be. In other words, the correlation performance index IPC gives an indication of how close the execution of the physical exercise is to a reference execution of the physical exercise.

For example, reference correlation coefficients are defined starting from the speed measures acquired by monitoring the execution of the exercise considered by at least one professional athlete. Advantageously, the server 3 stores a set of reference correlation coefficients for each type of physical exercise (without a main movement component) that can be analysed.

The correlation performance index IPC is then compared with a correlation performance threshold value (decision block 123), for example comprised between 80% and 90%, preferably equal to 88%.

In the event that the directional performance index IPD is equal to or greater than the directional performance threshold value (output branch Y of block 123), the physical exercise is considered to have been executed correctly, as in the case of Figure 7a, wherein the correlation coefficients r _xy , r _xz and r _yz are included in a neighbourhood of the reference correlation coefficients f _xy , f _xz and f _yz , forexample, tolerance ranges f _xy ± A _X y, r _xz ± D _cz and fy _Z ± Ay _Z , where A _xy , A _xz and A _yz . are values dependent on the selected directional performance threshold value. Also in this case, it is optionally envisaged transmitting a message of correct physical exercise to the fitness tracker 2 containing information confirming the correct execution of the physical exercise to the user U, for example comprising the value of the correlation performance index IPC (block 125).

Conversely, when the correlation performance index IPC is lower than the correlation performance threshold value (output branch N of block 123), the physical exercise is considered to have been executed incorrectly, as in the case of Figure 7b, wherein the correlation coefficients r _xy , r _xz and retail outside the tolerance ranges f _xy ± A _xy , f _xz ± A _xz and fyz ± Ayz . Accordingly, a message of incorrect physical exercise is transmitted to the fitness tracker 2 containing information that warns the user U of an incorrect execution of the physical exercise, for example comprising the value of the correlation performance index IPC (block 127).

Once the message of correct physical exercise (blocks 115 and 125) or the message of incorrect physical exercise (blocks 117 and 127) is transmitted, the system 1 continues to monitor the measures detected by the fitness tracker 2, in other words, the procedure returns to block 101 described above in order to identify the execution of other physical exercises.

However, it is clear that the above examples must not be interpreted in a limiting sense and the invention thus conceived is susceptible of numerous modifications and variations.

For example, in one embodiment the remote processing unit 3 is a user device - such as a personal computer, a smartphone and/or a tablet - configured to run an instance of the software application 4. Alternatively, it is possible to use a fitness tracker 2 with sufficient computing power to run an instance of the software application 4, thus omitting the remote processing unit 3 from the system 1 .

Again, nothing prohibits using a wearable electronic device other than the fitness tracker 2, for example a wearable device such as a chest strap.

In an alternative embodiment, the sensors may be configured to directly provide linear speed measures.

As will be apparent to the person skilled in the art, one or more steps of the above-described procedure may be performed in parallel with each other or in an order different from that presented above. Similarly, one or more optional steps may be added or removed from one or more of the above described procedures.

For example, in one embodiment, the machine learning algorithm, used to determine whether the user executes a physical exercise or not, is configured to determine the execution of an exercise by directly receiving the speed values as input. In other words, the machine learning algorithm is configured to identify the features of interest by analysing the speed values provided as input.

In one embodiment, the operation of identifying the execution of a physical exercise, the operation of determining whether the physical exercise comprises a main movement component, and the operation of determining the type of physical exercise are performed in a single step by means of a single Al algorithm or by a group of Al algorithms. In an alternative embodiment, the procedure firstly comprises identifying whether a speed v _x ^A , v _y ^A and v _z ^A has a value, e.g. an average value over an observation time, that is substantially greater than the value of the other speeds. For example, the value of the main speed v _x ^A , v _y ^A and v _z ^A is equal to or greater than 5 preferably 10 times the value of the other speeds, instead of identifying the exact type of physical exercise executed, before selecting which performance index IPD or IPC to calculate. In this case, the exact type of physical exercise is, preferably, determined only when the calculation of the correlation performance index IPC is selected in order to identify the correct set of reference correlation coefficients to be used in the calculation of the correlation performance index IPC.

In one embodiment, another indicator, such as a variance or other mathematical instrument representing the power or magnitude of the measures, is used in the calculation of the directional performance index instead of the dispersion index.

In addition or alternatively, the calculation of the correlation performance index can be based on correlation formulas between measures other than the Pearson correlation coefficient, obtaining results without requiring substantial modifications to the procedure.

In an alternative embodiment, a directionality error index IED is calculated instead of the directional performance index IPD. In particular, the directionality error index IED is an indication to what extent the physical exercise execution deviates from the ideal movement concentrated along the main direction of movement or in the plane of movement. In the preferred embodiment, the directionality error index IED, with respect to a physical exercise in which the main speed is the speed parallel to the axis ZA, is defined in percentage form as:

IED 100

In other words, when an exercise is considered to have been correctly executed, the speeds v _x ^A and v _y ^A developing in the plane a are contained in a tolerance radius n by the axis ZA, parallel to which main speed vA

In alternative embodiments, the reference thresholds of the performance indices can be configured manually, e.g. by the user or by a coach. Preferably, the numerical values of the reference thresholds can be based on the user experience (e.g., with different modifiers applied according to whether the user belongs to one of the following categories: expert, intermediate, beginner). In addition or alternatively, the numerical values of the reference thresholds may be based on an additional load - i.e. the weights - used by the user while executing the physical exercise: with heavier weights, the reference threshold must be more stringent, i.e. have a greater value, than an exercise executed with lighter weights, in which a less stringent threshold still ensures that the user does not risk injuries.

Optionally, the method comprises monitoring the directional performance index IPD or the correlation performance index IPC of a sequence of physical exercises executed consecutively in order to identify a decreasing trend in the value of the indices IPD or IPC as the number of executions of the physical exercise increases. When the decreasing trend exceeds a threshold value, the method comprises sending an alarm message to the fitness tracker 2 comprising an indication to the user to stop repeating the physical exercise. In other words, a user's fatigue can be detected at an early stage and he can be invited to stop the repetitions of the physical exercise at the appropriate time, thus avoiding fatigue-related injuries.

In addition to or as an alternative to the above-mentioned correct and incorrect physical exercise messages, a report message may be transmitted referring to the physical exercises executed during a physical exercise session. In such a case, the measures provided by the sensors are analysed to determine a termination of a training session - for example, if no physical exercise execution is identified within a predetermined time range since the last detected execution of a physical exercise. When the completion of the training session is determined, the report message is transmitted to the fitness tracker. Preferably, the report message contains an indication of the quality of execution of each set of exercises analysed and, optionally, one or more indications on which errors are made by the user during the execution of the physical exercises as well as any suggestions to correct these errors.

Finally, although the use of the speeds associated with the fitness tracker during the execution of the exercise is preferred, in other embodiments accelerations or information on the position of the fitness tracker over time, which can be derived from the acceleration measures, are used as components of movement to be analysed.

Naturally, all the details can be replaced with other technically-equivalent elements.

For example, as an alternative to or in addition to automatically identifying the start and/or the end of a training session, the software application can be configured to receive a message of the start, suspension and/or end of the training session provided by the user via the fitness tracker.

Furthermore, although reference has been made above to the identification of physical exercises comprising a main movement component along a single direction, alternatively or in addition it is possible to identify physical exercises comprising a main movement component lying on a plane -that is, whose main movement component comprises two speeds referring to two different reference axes - as in the case of kettlebell swings lying in the plane g illustrated in Figure 4.

In conclusion, the materials used, as well as the contingent shapes and dimensions of the aforementioned devices, apparatuses and terminals, may be any according to the specific implementation requirements without thereby abandoning the scope of protection of the following claims.