FIELD OF THE INVENTION

The invention relates to a method of acquiring information about at least one of cardiac activity and respiratory activity of a body of a vertebrate in physical contact with a surface of a supporting system. The invention further relates to a monitoring system for use in such a method.

BACKGROUND ART

Monitoring of a human being's vital signs is being used in, e.g., patient monitoring, sleep monitoring, monitoring of the health status of an individual, etc. However, for some of these applications, it is important that the monitoring be as little noticeable to the individual as possible. This is especially important if the monitoring is carried out for consumer applications, at home. In that case, vital sign monitoring must be particular unobtrusive, as convenient and as invisible as possible, such as to not interfere with the individual's routine, habits and preferences.

In the particular case of sleep monitoring, or sleep assessment, it is known that different stages of sleep and changes in certain physiological processes are interrelated. Examples of such physiological processes are cardiac activity, respiratory activity, body movement, etc. For more background, see, e.g., "Sleep staging using cardiorespiratory signals", S.J.Redmond et al., Somnologie 11 :245-256, 2007. If vital sign monitoring is being used within the context of sleep monitoring at home, the monitoring preferably fits in with the sleep environment of the individual being monitored. More specifically, the monitoring equipment is preferably installed so that it does not interfere with the individual's comfort and, ultimately, with the individual's sleep. Preferably, the monitoring should not require that sensors be attached to the individual's body, e.g., to the head of the individual, such as is currently the case in laboratories that study sleep processes.

Different approaches are known to the problem of unobtrusively monitoring vital signs. However, the known approaches have several drawbacks. A first drawback is that the monitoring is robust only to a specific modality, e.g., only to respiration or only to cardiac activity, often at the expense of reliably measuring the one or more other modalities. Other approaches suffer from degradation in the quality of the signals sensed by the sensor because a specific signal representative of a specific modality or of a specific physiological process emerges as mixed with signals representative of other modalities or of other physiological processes.

Ballistocardiography is a well-known technique for measuring certain aspects of cardiac activity and is based on sensing the ballistic forces (recoil and impact) on the body of the person being monitored. The ballistic forces are associated with cardiac contraction and ejection of blood, and with the deceleration of the blood flow. Consider a scenario wherein the person is lying on a substantially horizontal surface, e.g., the surface of a mattress. When the heart ejects blood, it does so mainly through the ascending aorta, in a longitudinal direction parallel to the person's spine and towards the person's head. When the heart takes in blood from the body, it also mainly does so in the same longitudinal direction. The heart's ejection and intake of blood are a result of the forces exerted by the heart on the blood, driving the blood through the circulatory system. According to Newton's third law, the forces exerted by the heart on the blood induce in the body reaction forces with a direction opposite to the direction of the forces exerted by the heart. It is assumed that these reaction forces do not affect the stationary position of the body relative to the mattress, but are exerted on the combined masses of the body and of the mattress. It is these reaction forces that are being sensed and recorded in ballistocardiography. The recorded signal is indicative of a vital sign in the 1 -20 Hz frequency range, and is monitored by noninvasive methods.

A first known method of measuring cardiac activity through ballistocardiography uses strain gauges as sensors. A strain gauge senses the strain at the surface of a structural element at which the strain gauge is mounted (e.g., glued). Subjecting the structural element to a mechanical load will cause the structural element to deform (elastically). The deformation is then sensed by the strain gauge.

A first example of using strain gauges in ballistocardiography is disclosed in US patent application publication 2008/0097250, titled "Biological Information Measuring Panel, and Biological Information Measuring Device", and incorporated herein by reference.

US2008/0097250 describes a set-up wherein strain gauges are mounted on a thin, flat plate placed on top of the mattress, directly under the body of the person. Because the plate does not have any fixed supports at its extremities, the absolute deflection caused by the person lying on the plate is relatively small and very much dependent on the mechanical properties of the mattress underneath the plate. The stiffer the mattress is, the smaller the deflection and therefore the less accurate will be the detecting of the vital signs. Furthermore, as the plate is positioned directly underneath the person's body, this set-up can become uncomfortable for the person lying on top of the plate.

A second example of using strain gauges in ballistocardiography is disclosed in US patent application publication 2010/0094139, titled "System and method for obtaining physiological data of a patient ^' ', assigned to Koninklijke Philips Electronics, and incorporated herein by reference. US2010/0094139 describes a set-up wherein strain gauges are mounted on slats which are parts of a mattress support system in a specific bed. This set-up overcomes the problem addressed in the discussion above of US2008/0097250, as the strain gauge is installed at a surface that is fixed at its extremities. As a consequence, each slat deflects according to the load on the bed. Since the extremities are fixed, this deflection is directly proportional to the load supported by the surface area of the slat. This will give a relatively accurate measurement of the load on that surface area, thus enabling a robust detection of vital signs. Additionally, the set-up of

US2010/0094139 solves the comfort issues addressed in the discussion above of

US2008/0097250, since the slats and strain gauges are positioned underneath the mattress.

However, the set-up in US2010/0094139 very much depends on the mechanical characteristics of the bed and, more importantly, of the mechanical characteristics of the mattress support in the bed's structure. Preferably, the set-up in US2010/0094139 is used with beds that have flexible slats that can be removed and later reinstalled after accommodating the strain gauges and wiring. Alternatively, add-on structures already equipped with strain gauges can be firmly attached to the slats already in place, thus avoiding that the slats need to be removed and/or the bed structure needs to be disassembled. The set-up of US2010/0094139 is especially attractive for the professional healthcare market, where only a limited number of types of beds are available. In the consumer market, however, the variety and types of beds available is much larger, rendering the set-up ofUS2010/0094139 less attractive.

A second known method of measuring cardiac activity through ballistocardiography uses one or more piezoelectric sensors. A piezoelectric sensor has a crystal that causes its electric charge to re-distributes when the crystal gets deformed under a mechanical load. This

piezoelectric effect can be used to measure, e.g., pressure, acceleration, strain or force. A piezoelectric sensor installed at a surface of a deformable object can be used to measure the amount of deformation of the object when experiencing a mechanical load. Consider a piezoelectric sensor that is mounted at the surface of a suitably shaped and suitably deformable object. Consider also a person lying substantially extended in a horizontal direction on a surface, e.g., a mattress. If the combination of the piezoelectric sensor and the object is then positioned underneath a region of the person's thorax, the changes in deformation of the object can be measured owing to cardiorespiratory activity of the person's body. The company EarlySense Ltd., markets the "EverOn" system that uses such a set-up to measure and record the vital signs of a patient such as heart rates, respiratory rates and movements. The "EverOn" sensor comprises a sensing plate with an integrated piezoelectric sensor placed between a mattress of a hospital- bed and the mattress support of the hospital-bed. The sensing plate has four short legs resting on the mattress support so as to prevent causing a bulge in the mattress that might interfere with the patient's comfort. The small size and thinness of the sensing plate render the "EverOn" sensor suitable for application with many different types of mass-market beds.

In general there is a trade-off between the size of a sensor (relative to the size of the bed, or relative to the size of the part of the bed occupied by the person) and the reliability in measuring respiratory activity. The smaller the width of the sensor, the higher the convenience and comfort to the person, but also, the more challenging it will be to reliably measure respiratory activity. More specifically, when a load with a relatively small footprint is applied at

approximately the center of the sensing plate, the load causes a deflection proportional to a magnitude of the downwards force being applied by the load. If the footprint of the load is larger than the width of the plate, then the downwards force will give rise to only a little deflection, or even no deflection at all, of the sensing plate. The fluctuating load induced by the movements of the patient's chest during respiratory activity has a relatively large footprint compared with the horizontal dimensions of the sensing plate. Depending on the position of the patient's chest relative to the sensing plate, depending on the size of the patient's chest and depending on the size of the sensing plate, it might well be that nearly no deflection will be measured and, as a result, that respiratory activity cannot be measured reliably. However, the ballistic effects caused by cardiac activity induce a load with a relatively small footprint so that cardiac activity can be measured reliably.

A third known method of measuring cardiac activity through ballistocardiography uses one or more acceleration sensors. An example thereof is discussed in "Praecordial

ballistocardiography", P. Mounsey, British Heart Journal, 1957, April, Vol.19 (2), pp 259-271. The acceleration sensor described in "Praecordial ballistocardiography" is an electrokinetic device that registers electrical effects produced bat a mercury-sulphuric acid interface. The device consists of a glass capillary tube containing alternate layers of metallic mercury and sulphuric acid, thereby creating multiple interfaces of the mercury and electrolyte solution. Movement of the tube produces changes at the surfaces of the mercury and electrolyte solution and these in turn give rise to variations in electrical potential that are proportional to the rate of change of velocity imparted to the tube. In comparison with the strain gauges and the piezoelectric sensors, discussed above, acceleration sensors have the advantage that they are relatively small. However, an acceleration sensor does not work if placed at the bed at stationary support that can be considered non-deformable in practical use for the purposes of measuring cardiac activity or respiratory activity. Because of their small size, one or more acceleration sensors could be installed between the patient and the mattress without causing a lot of discomfort, or between the pillow and the mattress. Alternatively, one or more acceleration sensors could be placed underneath the mattress, provided that the acceleration sensors are installed at a structure that is sensitive enough to get mechanically deformed and undergoes an excursion of adequate amplitude as a result of being subjected to the fluctuating load caused by the person's cardiac activity and/or respiratory activity.

The monitoring of cardiac activity and/or respiratory activity is addressed in the following publications, briefly discussed below.

US patent application publication 2009/0203972, incorporated herein by reference, discloses an apparatus, system, and method that monitor the motion, breathing, heart rate and sleep state of subjects, e.g., humans, in a convenient, non-invasive/non-contact, and low-cost fashion. More particularly, the motion, breathing, and heart rate signals are obtained through processing applied to a raw signal obtained in a non-contact fashion, typically using a radio- frequency sensor. Periods of sleep disturbed respiration, or central apnea can be detected through analysis of the respiratory signal. The mean heart rate, and derived information, such as the presence of cardiac arrhythmias can be determined from the cardiac signal. Motion estimates can be used to recognize disturbed sleep and periodic limb movements. The sleep state may be determined by applying a classifier model to the resulting streams of respiratory, cardiac and motion data. A means for display of the sleep state, respiratory, cardiac, and movement status may also be provided. US patent application publication 2010/0030085, incorporated herein by reference, addresses extraction of components of a signal captured by an accelerometer, obtaining information about physiological data such as cardiac, respiratory and snoring activity. The extracted signal components are useful for the diagnosis of different types of abnormal respiratory phenomena during sleep (apneas, hypopneas and respiratory efforts associated to micro-arousals).

US patent application publication 2009/02694784, incorporated herein by reference, discloses producing, by means of using a single sensor input, multiple different output signals for a polysomnograph (PSG) machine, the multiple different output signals including a first output indicative of an upper airway restriction (UAR), a second output indicative of an airway pressure during respiration, and a third output indicative of an airway air temperature during respiration.

The publication "Monitoring of Respiration and Heartbeat during Sleep using a Flexible Piezoelectric Film Sensor and Empirical Mode Decomposition", Nan Bu et al, Proceedings of the 29th Annual International Conference of the IEEE EMBS, Cite Internationale, Lyon, France, August 23-26, 2007, pp 1362 - 1366, addresses cardio-respiratory monitoring during sleep.

Cardio -respiratory monitoring is one of the basic means for assessment of personal health, and has been widely used in diagnosis of sleep disorders. The authors propose a method for noninvasive and unconstrained measurement of respiration and heartbeat during sleep. A flexible piezoelectric film sensor made of aluminum nitride (A1N) material is used in this study. This sensor measures pressure fluctuation due to respiration and heartbeat on the contact surface when a subject is lying on it. Since the A1N film sensor has good sensitivity, the pressure fluctuation measured can be further separated into signals corresponding to respiration and heartbeat, respectively. In the proposed method, the signal separation is achieved using an algorithm based on empirical mode decomposition (EMD). Experiments have been conducted with three subjects. The experimental results show that respiration and heartbeat signals can be successfully obtained with the proposed method.

The publication "Estimation of the Respiratory Component from Ballistocardiography Signal using Adaptive Interference Cancellation" , Xu Wang et al., Control and Decision

Conference (CCDC), May 23-25, 2011 Chinese, pp. 571 - 574, addresses real-time heart rate and respiration monitoring. Real-time heart rate and respiration monitoring is considered of major importance to be applied in the hospital and eventually at home. The authors propose a method for extracting respiratory signal from ballistocardiography, which is a completely non-invasive and unconstrained heart rate measurement method. The authors developed a system to detect ballistocardiography signal without subject's awareness. And then the adaptive interference cancellation algorithm, which is able to enhance the common information and attenuate the uncorrected noise, is applied to derive respiration component. The method was verified using about 1040 sets of 1-min long data collected from 20 healthy subjects. The ECG and nasal thermistor signals were recorded simultaneously as reference signal. The results showed that, compared with the reference data, the extracted pulse waveform was synchronized and the average error rate of quantitative assessment was feasible. The study shows that respiration can be derived under dynamic activities from ballistocardiography signal without significant differences from traditional methods.

SUMMARY OF THE INVENTION

As discussed above, vital signs can be monitored unobtrusively. The inventors have recognized as a drawback of the known approaches that the monitoring configuration may well prevent a robust measuring of respiratory activity and/or cardiac activity for several reasons.

A first reason is the issue of "signal masking". Signal masking occurs when the energy of a first signal stemming from monitoring a first modality (or: first physiological process) is significantly higher than the energy of a second signal stemming from monitoring a second modality (or: second physiological process) different from the first modality. In that case, the first signal masks or drowns the second signal, thereby making it harder to separate the second signal from the first signal. This may occur, for example, when sensors are being used that

simultaneously sense respiratory activity and cardiac activity in ballistocardiography. The location at which the sensor is installed relative to the bed, the properties of the structure supporting the sensor, and the position of the body lying in bed may all contribute to rendering the signal of a specific one of cardiac activity and respiratory activity dominant, masking the signal of the other one.

A second reason for preventing robust measuring is the issue of "spectral interference". Spectral interference occurs when the fundamental frequency of a periodic signal is close to the fundamental frequency or to a higher harmonic frequency of another signal. It might be difficult then to separate both signals. This is further aggravated by the nature of the environment wherein these sensors are deployed in operational use. For example, strain gauges or pressure sensors have been used under a mattress to measure forces induced by respiratory activity. If the mattress were a perfect medium for the propagation of pressure waves, there would be a linear dependence between the forces measured under the mattress and the forces the person's body exerts on the upper surface of the mattress by the periodic progressive inflation and deflation of the chest and/or abdomen. However, the materials used in the mattress, and the mattress's limited elasticity often cause the measured forces to be distorted, e.g., limited or truncated ("clipped"). As known, truncating of a signal is a non-linear transformation and thus introduces additional harmonics. For simplicity, assume that measuring cardiac activity through ballistocardiography gives rise to a signal that is a simple sine of a specific frequency (it is much more complicated, in practice). The frequency spectrum of the sine is a single peak at that specific frequency. When juxtaposed with the clipped respiratory signal, the peak of cardiac activity could easily be confused with a harmonics of the clipped signal representative of respiratory activity, thus making both signals virtually undistinguishable from each other in the frequency domain. In practice, the full system response is much more complex: noise severely distorts the signal as sensed, the mattress acts as mechanical filter with a non-linear phase-response and there is a wide variety of effects which depend on the body of the person, the position of the body relative to the bed, the location of the sensor relative to the body and to the bed, etc.

A third reason for preventing robust measuring is the issue of the "unpredictability of the set-up". The term "set-up" refers within this context to the physical characteristics of the actual configuration of the monitoring system as a whole that are relevant to the proper extraction of the signals representative of the vital signs. Examples of these characteristics include: the properties of the specific mattress actually used that are relevant to the propagation of the pressure waves induced by cardiac activity and/or of respiratory activity of the person's body, the configuration and the properties of the specific bed actually used, the characteristics of the body of the actual person that are relevant to its role as a source of pressure waves that are propagated through the mattress to the sensor, the tranquility of the environment that mechanically interacts with the bed (e.g., a floor that vibrates when a heavy truck is passing by or that vibrates as a result of loud music or of someone speaking with a booming voice, etc. The set-up is especially unpredictable in mass-market (e.g., consumer) applications. Due to the wide variety of mattresses, bed constructions, etc., it is very difficult to predict whether a particular sensor will sense vital signs of a single physiological process (e.g., only cardiac activity) or vital signs of multiple physiological processes (e.g., both of cardiac activity and respiratory activity. Moreover, in the case of using a single sensor to sense the vital signs of multiple physiological processes, it is difficult to predict whether each individual one of the multiple signals can be identified unambiguously.

The inventors propose a method of acquiring information about at least cardiac activity and respiratory activity of a body of a vertebrate in physical contact with a surface of a supporting system, e.g., including a bed and/or a mattress, or a seat, etc. The surface has substantially a pre-determined orientation relative to gravity. The method comprises generating a first signal by sampling a first magnitude of a first physical quantity. The first physical quantity is representative of a first interaction between the body and the surface due to a ballistic effect of the cardiac activity in a first direction substantially parallel to the surface. The method comprises generating a second signal by sampling a second magnitude of a second physical quantity. The second physical quantity is representative of a second interaction between the body and the surface brought about by a change in a pushing effect of the body on the surface in a second direction, substantially at right angles with the surface, due to an expanding or a contracting of the body as a result of the respiratory activity and due to the ballistic effect of the cardiac activity in the second direction. The method further comprises processing the first signal and the second signal for extracting at least one of a third signal representative of the ballistic effect of the cardiac activity in both the first direction and the second direction, and a fourth signal representative of the respiratory activity. The processing comprises using the first signal for identifying in the second signal a contribution from the ballistic effect of the cardiac activity in the second direction.

Consider the following example scenario, wherein the pre-determined orientation relative to gravity is a substantially horizontal orientation; the first direction is a substantially horizontal direction; and the second direction is a substantially vertical direction. The forces that the body exerts on the surface of the supporting system as a result of the ballistic effect of cardiac activity generally have horizontal vector components as well as vertical vector components if the body is lying on the surface that is oriented substantially horizontally. Respiratory activity gives rise to further forces on the substantially horizontal surface that in general have mainly vertical components. These forces and further forces exerted on the surface of the supporting system cause the surface of the supporting system to undergo elastic deformation. The elastic deformation initiated at the surface propagates through the supporting system towards one or more locations where the elastic deformations get sensed. The invention uses the facts that the first signal is mainly representative of the horizontal directionality of the ballistic effects of cardiac activity and that the first signal is correlated to the contribution of the ballistic effect of cardiac activity to the second signal in order to be able to identify in the second signal the contribution of cardiac activity as well as the other contribution of respiratory activity.

In more general terms, a process PA interacts with an environment and gives rise to a response RA from the environment; likewise, a process PB interacts with the environment and gives rise to a response RB from the environment. The response RA is vector with a first component RAl in a first dimension and a second component RA2 in a second dimension. The response RB is a vector that generally has only a second component RB2 in the second dimension. In the invention, the first component RAl is sensed and a combination of the second component RA2 and the second component RB2 is sensed. The characteristics of the first component RAl , e.g., its temporal or frequency characteristics, enable to disentangle and to determine the second component RA2 and the second component RB2 individually from the combination. Thus, the known difference in directionality of the responses RA and RB, together with the characteristics identified for the first component RAl enable to individually determine the vector components of the responses RA and RB.

In order to be able to identify the ballistic effect of cardiac activity and to identify the pushing effect of respiratory activity in the vital signs as captured, the first signal and the second signal are subjected to signal processing techniques, such as iterative analysis and adaptive filtering, as will be explained in detail later on. Accordingly, exploiting a difference in directionality between the ballistic effect and the respiratory effect enables to identify the individual vital signs in a robust manner.

The directional effects of the interaction of the body with the surface as a result of cardiac activity and of respiratory activity depend on the spatial configuration and orientation of the surface and on the orientation of the body with respect to the surface. For example, the ballistic effect of cardiac activity typically has a major horizontal component and a minor vertical component in case the body is lying on the surface that is oriented substantially horizontally (e.g., on a bed). The effects of respiratory activity, i.e., the effects of the cyclic expansion and contraction of the body's thorax and abdomen are mainly vertical in case the body is lying on the surface that is oriented substantially horizontally. As another example, the ballistic effect of cardiac activity typically has a major vertical component and a minor horizontal component in case the body is leaning against the surface that is oriented substantially vertically (e.g., the backrest of a chair or of a car seat). The effects of respiratory activity, i.e., the effects of the cyclic expansion and contraction of the body's thorax and abdomen are mainly horizontal in case the body is leaning against the surface that is oriented substantially horizontally. As yet another example, if the surface has a substantially horizontal part as well as a substantially vertical part, the effects of the interaction are a combination of those described in the preceding examples.

In an embodiment of the method in the invention, the generating of the first signal comprises sensing the first physical quantity at a first location at the supporting system, and the generating of the second signal comprises sensing the second physical quantity at a second location at the supporting system. The processing comprises subjecting at least one of the first signal and the second signal to time-shifting in order to adjust for a difference between a first delay present in the first signal and a second delay present in the second signal. The first delay results from a first effect of the first interaction propagating from the surface through the supporting structure to the first location. The second delay results from a second effect of the second interaction propagating from the surface through the supporting structure to the second location.

The body interacts with the surface of the supporting structure as a result of cardiac activity and respiratory activity. The interaction gives rise to elastic deformation of the surface that propagates further into the physical structure of the supporting system as, for example, longitudinal elastic waves (e.g., acoustic waves) and/or transverse elastic waves (e.g., shear waves). The propagation speed of an elastic wave through the physical structure is finite and depends, for example, the materials, mechanical properties and spatial configuration of the physical structure and, possibly, on the character of the waves (longitudinal, transverse, torsional, etc.). In first approximation it is assumed here that, in view of the relevant time scales of cardiac activity and of respiratory activity, propagation speed can be considered independent of the character of the waves. The body interacting with the surface of the supporting system can be considered as a spatial distribution of one or more wave sources at the surface. A difference between the first delay and the second delay will then depend on differences in lengths of the paths that the waves take through the supporting system between the positions of the one or more wave sources and the first location at which the first signal is being sensed, and the lengths of the paths that the waves take through the supporting system between the positions of the one or more wave sources and the second location where the second signal is being sensed. It is assumed here that the body can be considered as single wave sources at a particular position at the surface

In a further embodiment of a method according to the invention, the first physical quantity is an acceleration in the first direction, e.g., the horizontal direction, sensed at the supporting structure, and the second physical quantity is an acceleration in the second direction, e.g., the vertical direction, sensed at the supporting structure. The method comprises exploiting a multi- axial acceleration sensor for generating both the first signal and the second signal.

An advantage of using the multi-axial acceleration sensor is that the elastic deformation of the supporting system, due to the ballistic effect of cardiac activity in the first direction, and the other elastic deformation of the supporting system, due to a combination of respiratory activity in the second direction and the ballistic effect of cardiac activity in the second direction, propagate along the same path before reaching the multi-axial acceleration sensor. A further advantage is that only a single multi-axial acceleration sensor can be mounted at the supporting structure in order to configure the supporting system for implementing the method of the invention.

The invention can also be commercially exploited as a monitoring system configured for acquiring information about at least cardiac activity and respiratory activity of a body of a vertebrate, in physical contact with a surface of a supporting system. The surface has

substantially a pre-determined orientation relative to gravity. The monitoring system comprises a sensor system configured for generating a first signal by sampling a first magnitude of a first physical quantity, and for generating a second signal by sampling a second magnitude of a second physical quantity. The first physical quantity is representative of a first interaction between the body and the surface due to a ballistic effect of the cardiac activity in a first direction substantially parallel to the surface. The second physical quantity is representative of a second interaction between the body and the surface brought about by a change in a pushing effect of the body on the surface in a second direction, substantially at right angles with the surface, due to an expanding or a contracting of the body as a result of the respiratory activity and due to the ballistic effect of the cardiac activity in the second direction. The monitoring system comprises a signal processing system configured for processing the first signal and the second signal for extracting at least one of a third signal representative of the ballistic effect of the cardiac activity in both the first direction and the second direction, and a fourth signal representative of the respiratory activity, the processing comprises using the first signal for identifying in the second signal a contribution from the ballistic effect of the cardiac activity in the second direction.

Such a monitoring system comprising the sensor system and the signal processing system can be commercially exploited, e.g., as an after-market add-on to an existing bed or mattress. Alternatively, the monitoring system of the invention can be commercially exploited in a configuration of being physically integrated with the supporting system (e.g., installed on a bed or at a mattress or in a chair or seat) so that the monitoring system also includes the supporting system.

The pre-determined orientation relative to gravity is, for example, a substantially horizontal orientation, the first direction is then, e.g., a substantially horizontal direction, and the second direction is e.g., a substantially vertical direction.

In an embodiment of the monitoring system, the sensor system is operative to sense the first physical quantity at a first location at the supporting system for generating the first signal. The sensor system is operative to sense the second physical quantity at a second location at the supporting system for generating the second signal. The processing comprises subjecting at least one of the first signal and the second signal to time-shifting in order to adjust for a difference between a first delay present in the first signal and a second delay present in the second signal. The first delay results from a first effect of the first interaction propagating from the surface through the supporting structure to the first location. The second delay results from a second effect of the second interaction propagating from the surface through the supporting structure to the second location.

In another embodiment of the monitoring system, the first physical quantity is an acceleration in the first direction sensed at the supporting structure; the second physical quantity is an acceleration in the second direction sensed at the supporting structure, and the sensor subsystem comprises a multi-axial acceleration sensor for generating both the first signal and the second signal.

In a further embodiment the monitoring system, the monitoring system comprises the supporting system, e.g., a bed or a mattress. US patent application publication 2009/0203972, discussed in the background art section above, describes a method wherein a single raw signal from a single sensor is separated into three different physiological parameters: a cardiac signal, a respiratory signal and motion. Similarly, US patent application publication 2010/0030085, discussed above, describes the extraction from a single signal, captured by a single sensor, three signals, each respective one thereof

corresponding to a respective one of three different physiological parameters, to wit: cardiac activity, respiratory activity and snoring. Likewise, US patent application publication

2009/02694784, discussed in the background art section above, discloses producing multiple different output signals for a polysomnograph (PSG) machine from a single sensor input. The publication "Monitoring of Respiration and Heartbeat during Sleep using a Flexible

Piezoelectric Film Sensor and Empirical Mode Decomposition", Nan Bu et al, Proceedings of the 29th Annual International Conference of the IEEE EMBS, Cite Internationale, Lyon, France, August 23-26, 2007, pp 1362 - 1366, discussed in the background art section above, describes the use of one single sensor which outputs a single signal and describes a method to separate it further into different signals (in this case: respiration and heartbeat).

In contrast, the current invention addresses, among other things, a scenario wherein, due to the properties of the measuring medium or of the specific sensor used, the separation of a single signal into multiple signals is not possible, or at least, very difficult to achieve with acceptable accuracy. In order to solve this problem, the inventors propose introducing the additional capturing of only a single one of the physiological parameters. This approach facilitates the separation of the physiological parameters as captured.

The approach in publication "Estimation of the Respiratory Component from

Ballistocardiography Signal using Adaptive Interference Cancellation" , Xu Wang et al., Control and Decision Conference (CCDC), May 23-25, 2011 Chinese, pp. 571 - 574, discussed in the background art section above, differs slightly from the approach taken in the publication of Nan Bu et al. The approach in the publication of Xu Wang et al., seemingly uses two sensors (as indicated in their Fig. 1) to capture changes in pressure caused by the ballistic effects of the heart and the movements caused by respiration and muscle activity. Besides the fact that the use of two sensors installed under the chair is not described in the publication itself, it is clear that, by design, the two sensors give exactly the same information with respect to the physiological parameters that the sensors can capture. Namely, because of the locations of the two sensors, the two sensors will measure a mix of three physiological processes (cardiac, respiratory, body movements) and therefore offer no advantage in the separation of these signals. Again, the current invention introduces an additional sensor that captures only a single one of the physiological parameters, and not a mixture or combination of two or more physiological parameters. This approach facilitates the separation of the physiological parameters as captured by multiple sensors.

BRIEF DESCRIPTION OF THE DRAWING

The invention is explained in further detail, by way of example and with reference to the accompanying drawing, wherein:

Figs.l and 2 are diagrams illustrating measurable effects of cardiac activity of a human being;

Fig.3 is a diagram illustrating measurable effects of respiratory activity of the human being;

Fig.4 is a diagram with a graph of a (normalized) force or (normalized) pressure as a function of time, as a result of a person's respiratory activity;

Fig.5 is a diagram of the frequency spectrum of the graph of Fig.4;

Fig.6 is a diagram illustrating the effect of clipping on the graph of Fig.4;

Fig.7 is a diagram with the frequency spectrum of Fig.6;

Fig.8 is a diagram of a graph illustrating a highly simplified example of cardiac activity, showing only a fundamental frequency of a ballistocardiogram;

Fig.9 is a diagram with the frequency spectrum of the graph in Fig.8;

Fig.10 is a diagram of a graph resulting from combining the graphs of Figs. 6 and 8;

Fig.11 is a diagram of the frequency spectrum of the graph in Fig.10;

Figs.12, 13 and 14 are diagrams illustrating examples of a first embodiment of a set-up using a multi-axial acceleration sensor;

Fig.15 is a diagram illustrating an example of a second embodiment of a set-up using multiple single-axis acceleration sensors;

Fig.16 is a diagram illustrating an example of a third embodiment of a set-up, using two single-axis acceleration sensors; Fig.17 is a diagram illustrating an example of a fourth embodiment of a set-up, using a single-axis acceleration sensor and a piezoelectric sensor;

Fig.18 is a diagram of a highly simplified ballistocardiogram produced after filtering out the respiratory effects; and

Fig.19 is a diagram with the frequency spectrum of the ballistocardiogram of Fig.18.

Throughout the Figures, similar or corresponding features are indicated by same reference numerals.

DETAILED EMBODIMENTS

Figs.l and 2 are diagrams illustrating measurable effects of cardiac activity of a body 102 of a sleeping person lying substantially horizontally extended on a supporting system. The supporting system includes, e.g., a mattress 104, a bed (not shown separately) carrying the mattress 104, a floor (not shown separately) that supports the bed, etc. The sleeping person has a beating heart 106. Cardiac activity of the sleeping person exerts forces on the blood that in turn induce reaction forces on the body 102 in a direction 108 opposite to a direction of the forces exerted by the heart 106 on the blood. It is assumed that these reaction forces do not substantially affect the stationary position of the body 102 relative to the supporting system, and that these reaction forces are exerted on the combined masses of the body 102 and of the mattress 104 and the rest of the supporting system. It is the effect of these reaction forces that is being sensed and recorded in ballistocardiography. The reaction forces due to the ballistic effect have main vector components in a horizontal, or: longitudinal, direction 1 10. It would therefore seem advantageous to use a sensor 112, e.g., an acceleration sensor, which measured the longitudinal acceleration (i.e., along a direction 1 14 of the length of the body 102).

However, monitoring systems are known in the art that rely on measuring the effect of forces, representative of cardiac activity, that have a direction 202 orthogonal to the body 102. That is, the effect is being measured of forces, which have main vector components in a vertical direction 202, using an appropriate sensor 204. This is illustrated in the diagram of Fig.2. See, for example, US patent application publication 2008/0097250, US patent application publication 2010/0094139, and the "EverOn" system marketed by the company EarlySense Ltd., all discussed above. These known systems rely on the fact that the main arteries, which transport blood to the heart 106 or transport blood from the heart 106, do typically not run parallel to the horizontal direction 1 10. This is a result of, for example, an inclination in the posture of the body 102, as the person's head lies usually higher than the person's legs due to the use of a pillow 1 16, or due to deformations of the mattress 105 supporting the heavier part of the body 102 such as the person's chest. This implies that under typical circumstances the ballistic forces will have orthogonal components as well as longitudinal components. Accordingly, the forces due to the ballistic effect can, in principle, be measured along any or both of the orthogonal direction 202 and the longitudinal direction 1 10.

Fig.3 is a diagram illustrating measurable effects of respiratory activity of the body 102. Respiratory activity is manifested by means of the alternating inhaling and exhaling air via the body's nose and mouth, causing the person's lungs 302, and therefore the person's chest, to expand and contract alternately. As the person is lying horizontally on the supporting system in the gravitational field, the expansion of the person's chest causes the person's center of gravity to rise, and the subsequent contraction of the person's chest causes the person's center of gravity to drop. The upwards movement of the center of gravity is a result of the force exerted on the body 102 by the surface of the supporting system in response to the person's chest increasing its volume while gravity keeps pushing the body 102 against the supporting system. According to Newton's third law the supporting system in turn experiences a reaction force of equal magnitude, but in opposite direction. The downwards movement of the center of gravity is a result of the person's chest decreasing its volume. In the absence of gravity, the body 102 would retract from the surface of the supporting system. Owing to gravity, the body 102 keeps in contact with the surface of the supporting system but is momentarily pushing less hard on the surface while the center of gravity is being lowered. Accordingly, the supporting system experiences recurrent changes in the substantially vertical force of the body 102 on the surface in synchronism with the respiratory cycle of the person. These recurrent changes of the substantially vertical force at the surface cause recurrent elastic deformations of the surface, here of the mattress, that propagate through the supporting system and cause further recurrent elastic deformations elsewhere in the supporting system. A suitable sensor, e.g., a pressure sensor 304 located, for example, between the body 102 and the mattress or between the mattress and the rest of the supporting system, picks up these further recurrent elastic deformations that are representative of the person's respiratory activity. When monitoring cardio-respiratory activity, i.e., when monitoring cardiac activity as well as respiratory activity of a person, the above problem of signal masking has to be taken care of, owing to the fact that both cardiac activity and respiratory activity involve forces exerted on the supporting system, which have vertical components, as discussed above with reference to the diagrams of Figs.1 , 2 and 3. Furthermore, the problems of spectral interference and of the set-up unpredictability have to be considered as well.

As to the phenomenon of spectral interference, reference is made to the diagrams of Figs.

4-11.

Fig.4 is a simplified graph of a (normalized) force or (normalized) pressure as a function of time, as a result of a person's respiratory activity and determined through, for example, one or more strain gauges or pressure sensors at the supporting structure, e.g., located underneath the mattress. The mattress is, in this hypothetical case, a perfect medium for the propagation of elastic deformation, in the sense that the force, or pressure, measured underneath the mattress relates linearly to the actual forces caused by the progressive inflation and deflation of the person's chest and/or abdomen per individual respiratory cycle. If the person being monitored breathes regularly, the graph of Fig.4 is sinusoidal to a reasonable approximation.

Fig.5 represents the frequency spectrum of the graph of Fig.4, having a single peak at the frequency of the periodicity of the force or pressure, here 0.33 Hz.

However, the materials used in the mattress, combined with the materials' limited elasticity, often causes these forces to be limited or "clipped". This clipping effect is illustrated in the graph of Fig.6. The clipping distorts the originally sinusoidal character by subjecting the sinusoid to a non-linear transformation, thus introducing additional frequencies, as shown in the frequency spectrum of Fig.7.

Now consider cardiac activity, as represented in the highly simplified graph of Fig.8, that is indicative of the fundamental frequency of a hypothetical ballistocardiogram. The frequency spectrum of the graph of Fig.8 is shown in the graph of Fig.9. The frequency spectrum in Fig.9 shows a single peak at the ground frequency, here 1 Hz.

Assume now that a sensor is being used that picks up the effects of cardiac activity of the person as well as of respiratory activity of the person. The sensor then picks up a combination of the ballistic effect of cardiac activity and the clipped effect of respiratory activity. That is, if the graph of Fig.8 (forces due to ballistic effect of cardiac activity) and the graph of Fig.6 (clipped forces resulting from respiratory activity) are combined, the result is the graph of Fig.10, whose frequency spectrum is the graph of the diagram of Fig.l 1. Now, the peak in the frequency spectrum of the ballistocardiogram of Fig.9 could be easily be masked by one of the additional peaks that the clipping introduces into the frequency spectrum of respiratory activity as shown in the graph of Fig.7, thus making the contributions of cardiac activity and of respiratory activity virtually indistinguishable in the frequency domain.

In practice the full system response is much more complex: noise severely distorts the sensor signals, the mattress acts as a mechanical filter with a non-linear phase response and, of course, there is a wide variety of further effects which depend on the individual being monitored, his/her position in the bed, and the location(s) of the sensor(s).

As to set-up unpredictability, it is remarked here that, especially when developing solutions for the consumer market, it is difficult to predict in advance the properties of the set-up wherein the person is going to be monitored. Due to the variety of mattresses, bed constructions and personal characteristics, it is very difficult to predict whether a particular sensor will measure one or more physiological processes, e.g., one or more of cardiac activity and respiratory activity. Furthermore, in case a single sensor is used to monitor multiple physiological processes, it is difficult to predict which of the processes, if any, will give a dominant effect in comparison with the effect of one or more other processes.

Due to this complexity, the problem of achieving high-quality measurements of vital signs of different physiological processes using a single sensor or a sensor of a single type has not yet been solved satisfactorily.

Furthermore, owing to the nature of the known monitoring systems, especially when developed for the consumer market, it is sometimes impractical or even impossible to combine multiple sensors which, by construction or design, can robustly measure one single physiological parameter each. To give an example, if any of the sensors described above (strain gauges, piezoelectric sensors, acceleration sensors) is installed underneath the mattress (or between the mattress and the person to be monitored and lying on the mattress), such sensor will

simultaneously measure respiratory activity and cardiac activity. The extent, to which the effects of each of these physiological processes will be present in the sensor signal, heavily depends on the properties of the mattress, of the bed, of the bed frame and of the characteristics of the person lying on the bed. In the invention, the effects of the person's cardiac activity are sensed and first signals are generated that are fully or mainly attributed to the person's cardiac activity. The word "mainly" as used in the previous sentence signifies the fact that the influence of respiratory activity on the first signals is negligible compared to the influence of cardiac activity. In the invention, the combined effects of cardiac activity and respiratory activity are sensed as well and second signals are generated that are attributed to a combination of the person's cardiac activity and respiratory activity. The first signals and the second signals are then subjected to signal processing techniques in order to extract a third signal representative of the ballistic effect of cardiac activity in both the horizontal direction and the vertical direction, and a fourth signal representative of respiratory activity. Below, various ways are being discussed to set up a monitoring system using the approach of the invention for producing the third signal and the fourth signal.

Figs.12, 13 and 14 are diagrams to illustrate examples of a first embodiment 1200 of a setup wherein a multi-axial acceleration sensor 1202 is used for measuring the horizontal effect owing to the horizontal vector component of the ballistic effect representative of cardiac activity, and for measuring the vertical effect owing to the combination of the vertical vector component of the ballistic effect representative of cardiac activity and the vertical vector component of the forces generated by respiratory activity. In Fig.12, the multi-axial acceleration sensor 1202 is shown as positioned between the mattress 1204 and the rest (not shown) of the supporting system. In Fig.13, the multi-axial acceleration sensor 1202 is positioned between the body 102 and the mattress 1204. In Fig.14, the multi-axial acceleration sensor 1202 is positioned between a pillow 1402 and the mattress 1204.

Measuring respiratory activity at this location requires extremely sensitive accelerometers, which are able to pick up the small vertical deformations at this location caused by the large load- footprint of an inflating (and deflating) thorax. The proximity to the thorax and the fact that, due to the type of materials used in mattresses, deformations are not restricted to the exact area where pressure is being exerted, makes this one of the locations where with proper technology respiration can still be measured. Although technically more difficult to execute, the inventors have experimentally determined such approach is feasible, provided that accelerometers are being used that have high sensitivity, low signal-to-noise ratio, and an appropriate natural frequency ("eigenfrequency"). The locations indicated in Figs.12, 13 and 14 are just a few of the large number of possibilities, although it is easy to imagine that different combinations of locations and sensing modalities could be used just as effectively.

Fig.15 is a diagram to illustrate an example of a second embodiment 1500 of a set-up, wherein multiple single-axis acceleration sensors, e.g., a first acceleration sensor 1502 and a second acceleration sensor 1504, are being employed so as to sense accelerations (a vector quantity) in different directions. For example, the first acceleration sensor 1502 is located on a sidewall of the mattress 1204 and is configured for sensing horizontal accelerations

representative of the ballistic effects of cardiac activity. The second acceleration sensor 1504 is located between the pillow 1402 and the mattress 1204 and is configured for sensing vertical accelerations representative of a combination of the ballistic effects of cardiac activity and respiratory activity.

The combination of the first acceleration sensor 1502 and the second acceleration sensor 1504 senses the effects of cardio -respiratory activity, as well of environmental noise caused by, e.g., a train or truck passing by, another person walking on the boards forming the floor supporting the bed, etc. The first acceleration sensor 1502 and the second acceleration sensor

1504 are located a certain distance apart from each other. This distance is typically much shorter than another distance between any specific one of the first acceleration sensor 1502 and the second acceleration sensor 1504 and a source of environmental noise. As a result, the effects of the environmental noise will be very similar at the locations of the first acceleration sensor 1502 and the second acceleration sensor 1504. At least, the effect of the environmental noise at the location of the first acceleration sensor 1502 and the effect of the environmental noise at the location of the second acceleration sensor 150 are strongly correlated, so that these effects can be eliminated through suitable processing of the sensor signals.

If the environmental noise effectively masks the effects of the physiological processes to be sensed, then suitable techniques can be exploited which rely on the fact that the physiological processes are being sensed by both sensors.

Algorithms can be used which exploit common information in two signals (e.g. blind source separation). A slightly more intricate procedure of processing of the signals is then needed. For example, one could apply the following algorithm under the following assumptions .A sensor signal SI from a sensor 1 is a combination of the effect of a first physiological process PI and the effect of environmental noise El . A sensor signal S2 from a sensor 2 is a combination of the effect of the first physiological process PI and the effect of a second physiological process P2. In order to determine the effect of the first physiological process PI and the effect of the second physiological process P2 individually, an algorithm is used which exploits the fact that sensor signal S 1 and sensor signal S2 have information in common about the first physiological process PI , so as to determine an estimate of the effect of the first physiological process PI by analyzing sensor signal S 1 and sensor signal S2. Then, the estimate of the effect of the first physiological process PI is used to determine an estimate of the effect of the second

physiological process P2 in the sensor signal S2, e.g., using the approach described earlier with reference to cardiorespiratory activity. Please note that the more intricate procedure is preferably used only in case there would be masking or spectral interference between the effect of the environmental noise El and the effect of the first physiological process PI . If there is neither masking nor spectral interference, a simple technique such as filtering, e.g., band-pass filtering, could be applied to identify the effect of the first physiological process PI in the sensor signal S 1.

Alternatively, a third sensor is used, placed at a location at or near the supporting structure, where the effects of the first physiological process P 1 and of the second physiological process P2 are negligible, and where only the effects of the environment noise El are picked up. The noise can be removed from the sensor signals SI and S2 which do sense the effects of the physiological processes. The use of such a third sensor is described in US patent application publication 2010/0016685 "Ballistocardiographic sensor system with a sensor arrangement and method of ballistocardiographic detection of body movements ^' ", assigned to Koninklijke Philips Electronics and incorporated herein by reference.

As another alternative, the effects of the first physiological process PI and the second physiological process P2 can be identified by means of combining the iterative procedure, described below, with adaptive filtering in the following way, wherein it is assumed that the effect of the environmental noise El in the sensor signal S 1 and the effect of the second physiological process P2 in the sensor signal S2 have different spectral properties. The sensor signal SI is used to design a filter, e.g., a band-pass filter, that is then applied to the sensor signal S2. As the effect of the environmental noise El is not present in the sensor signal S2, the filtering of the sensor signal S2 produces an estimate of the effect of the first physiological process PI . The estimate of the effect of the first physiological process PI is then, in turn, used to design another filter, e.g., a band-stop filter, which is applied to the sensor signal S2. As the sensor signal S2 is a combination of the effect of the first physiological process PI and the effect of the second physiological process P2, the filtering produces an estimate of the effect of the second physiological process P2

Consider the case of using multiple sensors, e.g., multiple acceleration sensors. Each respective one of the sensors senses the effects of cardiac activity and/or of respiratory activity after a respective propagation delay. The effects propagate from the surface of the supporting structure, which is in contact with the body 102, to the respective sensor via a respective propagation path through the supporting structure, and the propagation takes time. If the propagation paths have substantially different lengths, e.g., as a result of a substantial distance between the respective locations of the respective sensors, the signals supplied by the sensors are preferably pre-processed so as to take into account the difference in propagation delays. This will be discussed in more detail further below.

Fig.16 is a diagram to illustrate an example of a third embodiment 1600 of a set-up, using a third acceleration sensor 1602 and a fourth acceleration sensor 1604. The third acceleration sensor 1602 is placed between the body 102 and the mattress 1204 in a first position where both cardiac activity and respiratory activity can be sensed. The first position is, for example, within the region of the mattress 1204 in contact with the person's chest area. The fourth acceleration sensor 1604 is placed between the body 102 and the mattress 1204 in a second position where only, or mainly, the ballistic effects of cardiac activity can be sensed, such as in the region of the mattress 1204 in contact with the person's hips.

Fig.17 is a diagram to illustrate an example of a fourth embodiment 1700 of a set-up, wherein multiple sensors are being used, each different one thereof sensing a different physical quantity. For example, the fourth embodiment 1700 comprises a fifth acceleration sensor 1702 configured for sensing acceleration in a longitudinal direction (i.e., a horizontal direction along the length of the mattress 1204), and a piezoelectric sensor 1704 for sensing pressure exerted on a horizontally mounted deformable plate f the piezoelectric sensor 1704. The fourth embodiment 1700 has an advantage in that its spatial configuration is compact and simplifies collection, processing and separation of the signals from the fifth acceleration sensor 1702 and from the piezoelectric sensor 1704, as the fifth acceleration sensor 1702 and the piezoelectric sensor 1704 are positioned close together or are even accommodated together within the same housing of a physical device. As discussed above, an interesting aspect of the invention resides in providing a method of acquiring information about at least cardiac activity and respiratory activity of the body 102, which lies substantially extended in a horizontal direction on a surface of a supporting system, e.g., the upper surface of the mattress 1204. In the invention, a first signal is generated by sampling a first magnitude of a first physical quantity (e.g., a scalar or a vector) that is representative of a first interaction between the body 102 and the surface due to a ballistic effect of cardiac activity in the horizontal direction. A second signal is generated by sampling a second magnitude of a second physical quantity. The second physical quantity is representative of a second interaction between the body 102 and the surface. The second interaction is brought about by a change in a pushing effect of the body on the surface in the vertical direction due to an expanding or a contracting of the body 102 when inhaling or exhaling, and due to the ballistic effect of cardiac activity. The first signal and the second signal are then processed for extracting at least one of a third signal and a fourth signal. The third signal is representative of the ballistic effect of cardiac activity in both the horizontal direction and the vertical direction. The fourth signal is representative of respiratory activity.

The first signal and the second signal can be processed in a variety of manners to extract at least one of the third signal and the fourth signal.

A first manner involves signal subtraction. Signal subtraction can be applied if the first physical quantity and the second physical quantity have the same physical dimensions and the first signal and the second signal are expressed in the same physical units, e.g., the SI units.

As to the criterion of the first physical quantity and the second physical quantity having the same physical dimensions, consider the following example scenarios. In a first example, the first physical quantity and the second physical quantity are being sensed using a single multidirectional acceleration sensor, or the first physical quantity is sensed using one or more first single-direction acceleration sensors and the second physical quantity is sensed using one or more second single-direction acceleration sensors of similar design as the first single-direction acceleration sensors. Then, the first physical quantity represents acceleration (having dimensions expressed in SI units as m/sec ) in the horizontal direction and the second physical quantity represents also acceleration (expressed in SI units as m/sec ) in the vertical direction. In a second example, the first physical quantity is being sensed via one or more first strain gauges at the supporting structure and oriented in a particular orientation, and the second physical quantity is being sensed via one or more second strain gauges at the supporting structure and oriented in another orientation, the first strain gauges and the second strain gauges being of similar design. As known, a strain gauge measured the deformation of a physical object at a surface of the physical object. In operational use, a strain gauge forms part of an electric circuit. Deformation of the physical object causes the electrical resistance of the strain gauge to change, which in turn causes a voltage change in the electric circuit that is measured. The first physical quantity and the second physical quantity may then both have the dimension of volts (V). In a third example, the first physical quantity is being sensed via one or more first piezoelectric sensors at the supporting structure and oriented in a particular orientation, and the second physical quantity is being sensed via one or more second piezoelectric sensors at the supporting structure and oriented in another orientation, the first piezoelectric sensors and the second piezoelectric sensors being of similar design. As known, a piezoelectric sensor uses the piezoelectric effect to measure pressure, acceleration, strain or force (referred to as "mechanical load"). The piezoelectric sensor has a piezoelectric crystal that deforms when subjected to a mechanical load. The deformation of the piezoelectric crystal gives rise to a redistribution of an electrical charge through the crystal which can be measured as a voltage.

In the above, the sensors for sensing the first physical quantity and the sensors for sensing the second physical quantity have been specified as being of similar design. A similar design implies that the sensors have similar transfer functions for converting an input to an output, so that each of the sensors provides substantially the same specific output given a specific input.

Now, assume that the first physical quantity and the second physical quantity have the same physical dimensions and are being sensed by sensors of similar design. The first signal is representative of only cardiac activity as the relevant sensor senses the horizontal ballistic effects. The second signal is representative of a combination of the vertical ballistic effects of cardiac activity and the changes in the pushing down of the body 102 on the surface as a result of respiratory activity. Assume further that a relationship exists between the first signal (horizontal ballistic effects) and the component in the second signal represents the vertical ballistic effect, and that this relationship is known. For example, this relationship is a fixed ratio. Then, subtracting the first signal from the second signal (or the second signal from the first signal) after proper scaling using this relationship, produces a fourth signal substantially representative of only respiratory activity. The third signal that is representative of cardiac activity is then determined by the first signal and the known relationship. For a specific case, wherein one or more acceleration sensors are dedicated to sensing horizontal acceleration and one or more acceleration sensors are dedicated to sensing vertical acceleration, subtracting the magnitude of the horizontal acceleration from the magnitude of the vertical acceleration (or the other way round) provides a better estimation of effects of respiratory activity.

Signal subtraction can also be used if a specific one of the first physical quantity and the second physical quantity has physical dimensions equal to the time derivative of the other one of the first physical quantity and the second physical quantity. Then, sensing the specific one of the first physical quantity and the second physical quantity produces a specific one of the first signal and the second signal. Integration over time of the specific signal generates an intermediate signal representative of the time-integrated version of the specific one of the first physical quantity and the second physical quantity. The time -integrated version of the specific one of the first physical quantity and the second physical quantity is dimensionally compatible with the other one of the first physical quantity and the second physical quantity. Assume that the relationship is known between ballistic effects as represented in the intermediate signal and as represented in the other one of the first signal and the second signal. Then, subtracting the intermediate signal from the other one of the first signal and the second signal, and using proper scaling based on the known relationship, provides a better estimation of effects of respiratory activity.

A second manner of processing the first signal and the second signal to extract at least one of the third signal and the fourth signal is based on iterative analysis. In iterative analysis, results of analyzing the first signal (e.g., in the time domain or in the frequency domain) are used to process the second signal (e.g., by bandstop filtering on the frequency band of the first signal to remove from the second signal the vertical ballistic effects) and then back to improve the first signal again. This approach will be particularly effective if the analysis is done in the frequency domain and/or if the first physical quantity and the second physical quantity are not related by simple time-integration or time-derivation.

A third manner of processing the first signal and the second signal to extract at least one of the third signal and the fourth signal is based on adaptive filtering. In adaptive filtering, filter coefficients are determined based on the spectral response of the first signal (representative of the horizontal ballistic effects). These filter coefficients are then used to band-pass filter the second signal (representative of a combination of the vertical ballistic effects due to cardiac activity and of the effects of respiratory activity) for obtaining a better estimation of cardiac activity or to band-stop filter the second signal for obtaining a better estimation of respiratory activity.

To illustrate the adaptive filtering, consider the graphs of Figs.8 and 9. Fig.8 is indicative of a ballistocardiogram that corresponds to a simple sinusoid. The frequency spectrum of the graph of Fig.8 is shown in the graph of Fig.9. The frequency spectrum in Fig.9 shows a single peak at the ground frequency, here 1 Hz. The spectrum of the first signal (horizontal ballistic effects due to cardiac activity) is used for a least-square linear-phase FIR filter design of order 100, and this filter is then applied to the second signal represented in the graph of Fig.10. The graph of Fig.10 is the result of combining the graph of Fig.8 (forces due to vertical ballistic effect of cardiac activity) and the graph of Fig.6 (clipped vertical forces resulting from respiratory activity). Thus filtering out the second signal produces the graph of Fig.18. Note that in the graph of Fig.18 the ballistocardiogram is perfectly reconstructed upon an initial transient and that the influence of the clipped respiratory signal is removed from the second signal. Fig.19 is a diagram of the frequency spectrum of the graph in Fig.18. Adaptive filtering within the context of the invention has the advantage that it can be used independently of the nature of the first signal and the second signal, apart from the fact that characteristics of the first signal are also represented in the second signal.

Consider a scenario wherein multiple sensors are being used positioned at various locations at the supporting structure. One or more of the sensors sense the horizontal ballistic effects of cardiac activity only, and one or more other ones of the sensors sense a combination of the changing vertical pushing effect of respiratory activity and the vertical ballistic effects of cardiac activity. Examples of such scenario have been illustrated in the diagrams of Fig.15, 16 and 17. As explained above, there is a correlation between a first sensor signal representative of the horizontal ballistic effects and the component in the second sensor signal representative of the combination of the changing vertical pushing effect of respiratory activity and the vertical ballistic effects of cardiac activity. And this correlation is then used in the processing of the first sensor signal and the second sensor signal in order to determine individually the effect of cardiac activity and the effect of respiratory activity. The first sensor signal and the second sensor signal as subjected to the processing are assumed to be expressed on the same time base. That is, the correlation is assumed to be valid with respect to the first sensor signal captured at a particular moment and the second sensor signal captured at the same particular moment. As explained earlier, the elastic deformation, resulting from the interaction of the body 102 with the horizontal surface of the mattress 1204 owing to cardiac activity and respiratory activity, propagate through the supporting structure and are then sensed at various locations. It may then well be that the effect of the elastic deformation at the surface incurs different propagation delays before being sensed at the various locations owing to, e.g., different lengths of the propagation paths. For example, in the set-up illustrated in the diagram of Fig.16 and discussed above, the third acceleration sensor 1602 is located between the body 102 and the mattress 1204 within the region of the mattress 1204 in contact with the person's chest area, where both cardiac activity and respiratory activity can be sensed. The fourth acceleration sensor 1604 is located between the body 102 and the mattress 1204 in the region of the mattress 1204 in contact with the person's hips, so as to only, or mainly, sense the ballistic effects of cardiac activity. The third acceleration sensor 1602 is closer to the person's chest than the fourth acceleration sensor 1604. As a result, the effects of cardiac activity sensed by the third acceleration sensor 1602 at a certain moment will be sensed by the fourth acceleration sensor 1604 after a time delay as a result of difference in length of a propagation path between the chest and the third acceleration sensor 1602 and the length of another propagation path between the chest and the fourth acceleration sensor 1604. Signal processing methods are known to estimate this time delay and to adjust the sensor signals for this time delay see, for example, "A Survey and Comparison of Time-Delay Estimation Methods in Linear Systems ", S. Bjorklund , Linkoping Studies in Science and Technology, Thesis No. 1061 , Division of Automatic Control, Department of Electrical Engineering,

Linkopings universitet, 2003.

For completeness, it is remarked here that in addition to, or as an alternative to, acceleration sensors other sensors can be used that have other sensing modalities. For example, the inventors have found that a robust configuration is obtained by means of sensing the horizontal ballistic effect of cardiac activity, of a person lying horizontally on the surface via one or more accelerometers, and the vertical effects of cardiac activity and of respiratory activity via a piezoelectric sensor sensing a deformation of a piezoelectric element. The effects of cardiac activity and of respiratory activity can also be sensed by means of position sensors (e.g., via a piezoelectric sensor or an optical sensor), or by means of velocity (i.e., direction and magnitude of a change in position per unit time), or by means of registering the Doppler effect, etc. Note that acceleration can be determined from the position as a function of time by taking the second time derivative, and from the velocity as a function of time by means of taking the first time derivative.