SYSTEM AND METHOD FOR PREDICTING AN INDIVIDUAL'S FALL

Abstract

SYSTEM AND METHOD FOR PREDICTING AN INDIVIDUAL'S FALL The invention relates to a system and method for predicting an individual's fall, comprising: a casing (10) worn by an individual (8) in which is housed a plurality of sensors for acquiring measurements representative of the individual's posture and / or movements; a processing unit (100) for the measurements provided by said plurality of sensors and comprising: a module (110) for determining a profile score, based on data representative of the individual's specific characteristics; a module (120) for calculating motor cursors, based on the measurements provided by said plurality of sensors, and representative of the individual's activity over a predetermined time interval; a module (130) for calculating a fall risk score, based on said motor cursors and said profile score;a module (140) for determining a fall risk, a function of a variation of said fall risk score beyond a predetermined threshold over a predetermined time interval. Figure for the abstract: figure 1;

Description

Title of the invention: SYSTEM AND METHOD FOR PREDICTING AN INDIVIDUAL'S FALL Technical field of the invention

[0001] The invention relates to a system and method for predicting an individual's fall, that is, a system and method configured to assess and anticipate an individual's risk of falling. The invention also relates to a system and method for detecting an individual's fall, implementing a fall prediction system and method according to the invention. Technological background

[0002] Today there are many connected devices designed to detect falls of people likely to fall such as elderly people, people with attention deficit disorder, people prone to panic attacks, people with epilepsy, etc.

[0003] These devices and methods generally employ motion sensors (such as accelerometers) worn by the individuals being monitored, combined with data processing units for the data provided by these sensors. The processing unit analyzes and interprets the data values ​​provided by the sensors and triggers a fall detection alert as soon as the values ​​exceed a predetermined threshold. The proposed processing methods consist, for example, of counting the number of steps taken by the monitored individual over a predetermined period and triggering the alert if no activity is observed over a normal activity period.

[0004] These devices are now available in the form of pedometers, smartwatches, necklaces, smartphones, etc., which are worn by the individuals to be monitored. The processing units can be embedded in the devices or hosted on remote servers connected to the detection devices by wireless means.

[0005] These solutions are useful for being able to react quickly as soon as a fall is detected by the system, which makes it possible to quickly contact the monitored person and / or send emergency services to the scene for rapid assistance.

[0006] That being said, these systems do not prevent falls. In other words, the known solutions simply detect falls and reduce the intervention time between the fall and the care of the individual who has fallen.

[0007] The inventors sought to go further and to implement a system which makes it possible to predict the fall, before it occurs, so as to be able to avoid it. In other words, the inventors sought to develop a system and a method for predicting the fall of an individual, that is to say a system and a method which makes it possible to detect the imminence of a fall of a monitored individual, so as to be able to alert the individual (or a third party helping) that he risks falling, and thus to avoid the occurrence of the fall or to reduce its consequences. Objectives of the invention

[0008] The invention aims to provide a system and a method for predicting the risk of an individual falling.

[0009] The invention also aims to provide, in at least one embodiment, a system and a method that adapt to the physical and / or physiological characteristics of the person being monitored.

[0010] The invention also aims to provide, in at least one embodiment, a system and a method which can interact with the monitored individual in order to alert him or her to a risk of falling (or a third party helping him or her) and / or confirm a possible fall detected.

[0011] The invention also aims to provide, in at least one embodiment, a system and a method which makes it possible to monitor activity parameters of a monitored individual.

[0012] The invention also aims to provide a system and a method for detecting a fall of an individual which implements a system and a method for predicting a risk of a fall of an individual according to the invention. Description of the invention

[0013] To this end, the invention relates to a fall prediction system for an individual comprising: - a device configured to be worn by an individual, said device comprising a plurality of sensors for acquiring measurements representative of the posture and / or movements of the individual, including at least one accelerometer, one magnetometer and one gyroscope, - a unit for processing the measurements provided by said plurality of sensors.

[0014] The system according to the invention is characterized in that said processing unit comprises at least: - a module for determining a piece of data, called a profile score, which is a function of data representing the individual's specific characteristics, - a data calculation module, using motor cursors, based on measurements provided by said plurality of sensors, and representative of the individual's activity over a predetermined time interval, - a module for calculating a piece of data, called the fall risk score, based on said motor sliders and said profile score, - a module for determining fall risk, based on a variation said risk score of falling beyond a predetermined threshold over a predetermined time interval.

[0015] The system according to the invention is therefore unique in that it processes data acquired by sensors housed in a device worn by the monitored individual to deduce a fall risk based on the variation of a risk score beyond a predetermined threshold. The fall risk score is derived from motor indicators and a profile score specific to the individual. The motor indicators reflect the activity of the monitored person and provide a representation of the monitored person's physiological factors.

[0016] In other words, the system according to the invention continuously monitors the fall risk score and triggers an alert as soon as this risk score varies rapidly over a predetermined period of time, which is characteristic of a significant degradation in the stability of the person being monitored.

[0017] Throughout the text, the terms "supervised person", "person to be supervised", "monitored individual" or "individual to be monitored" refer to the person likely to fall who uses the system according to the invention.

[0018] Fall prediction is specific to each individual insofar as it relies on a profile score, which depends on the individual's own characteristics, and motor indicators, which are functions of the person's activity. Thus, and unlike most known systems, the invention adapts to the specific characteristics of the individual.

[0019] Advantageously and according to the invention, said module for determining said profile score of said individual comprises an automatic calculation model trained to determine a profile score, this calculation model, said first model, having been trained by means of a training database, said profile database, which includes data representative of characteristics specific to a plurality of individuals associated with occurrences of falls detected by said plurality of individuals.

[0020] According to this advantageous variant, the determination of an individual's profile score is based on an automatic calculation model trained using a profile database. This profile database consists of the specific characteristics of a plurality of individuals associated with detected occurrences of falls in this plurality of individuals.

[0021] The use of such an artificial intelligence model makes it possible to automatically detect characteristic patterns of fall risk.

[0022] The automatic calculation model implemented by the module for determining an individual's profile score can be of any type. It can be a supervised learning neural network, a support vector machine (more commonly known under the English acronym SVM for Support Vector Machine) or any other "machine learning" algorithm.

[0023] This module therefore makes it possible to assign to each individual equipped with the system according to the invention a profile score which depends on his or her own characteristics.

[0024] Advantageously and according to the invention, the data representing the specific characteristics of each individual include one or more pieces of information related to the age, sex, prescribed medications, weight, height, sight, hearing, history of falls, etc. of said individual.

[0025] According to this advantageous variant, the individual characteristics used to determine a profile score are information related to his age, sex, prescribed medications, weight, height, vision, hearing, history of falls, etc.

[0026] A specific weight can be used for each characteristic. For example, the individual's age can be assigned a weight of 90%, their sex a weight of 70%, the number of daily medications prescribed a weight of 75%, wearing glasses a weight of 50%, a history of falls a weight of 100%, the individual's weight a weight of 50%, and the individual's height a weight of 25%. These weights characterize the importance of the criterion in determining the profile score.

[0027] Of course, the weights indicated are only an example of the relevance given to the different criteria in determining the profile score and a different allocation of the weights of each criterion can be made without calling into question the principle of the invention.

[0028] Once the individual's profile score is known, the system according to the invention can calculate the individual's motor cursors over time. These motor cursors are intended to represent the individual's activity and behavior over predetermined time periods.

[0029] Advantageously and according to the invention, said motor cursors calculated by said calculation module are chosen from the group comprising: - sliders representing the average activity of the individual over a predetermined time interval T, called SMA sliders, calculated from the measurements provided by each of the sensors housed in said casing according to the following formulas: • [Math 1]

[0030] [Math.l] / rt+T rt+T pit+7" \ Or [Math.l] af, a?, a%

[0031] represent the acceleration values ​​measured on three axes of a right-handed trihedron (x,y,z) of said accelerometer, • [Math 2]

[0032] [Math.2] \m}dt4-+ Or [Math.2] mf, represent the magnetic values ​​measured on three axes of said right-handed trihedron (x, y, z) of the magnetometer, • [Math 3]

[0033] [Math.3] / çt+T rt+T \ = \g*\dt+\t Itffl^j Or [Math.3] represent the gyroscopic values ​​measured on three axes of said right-handed trihedron (x, y, z) of the gyroscope, - sliders representing the individual's instantaneous activity calculated from measurements provided by each of the sensors housed in said casing according to the following formulas: • [Math 4]

[0034] [Math.4] || A || t ^af + af + af •> • [Math 5]

[0035] [Math.5] || M || ( = 2 + m, 2 +mf 2 • [Math 6]

[0036] [Math.6] It G || ^gf + gf + gf sliders representing the energy of said individual, called HA sliders, calculated as the variation of instantaneous activity over said predetermined period T according to the following formulas: • [Math 7]

[0037]

[0038]

[0039] [Math.7] HA? = var[ || A || J where var represents the variation of the value over the predetermined time interval T, • [Math 8] [Math. 8] HA^ = var( || M || J where var represents the variation of the value over the predetermined time interval T, • [Math 9] [Math.9] HA% = var( || G || J where var represents the variation of the value over the predetermined time interval T, - sliders representing the harmony of the individual's activity, called HM cursors, over the said predetermined time interval according to the following formulas: • [Math 10]

[0040] [Math. 10] Or

[0041]

[0042] [Math. 10] HA „ A||A||t II 21 f “ A / • [Math 11] [Math. 11] = EEü T vHiimiij where [Math. 11] It M ||f = ^ • [Math 12] [Math. 12] rjmg_ Mimü tllvl'j- — il----7---*---T* OR [Math. 12] He ? = "V" - sliders representing irregularities in the frequency domain during the individual's activities over the predetermined time interval T, called HC sliders, measured according to the following formulas: • [Math 13]

[0043] [Math. 13] Or [Math. 13] h 4" 11 IM 11,=-5-1 • [Math 14]

[0044] [Math. 14] Or [Math. 14] • [Math 15]

[0045] [Math. 15] Or [Math. 15] It G ||, = ^1

[0046] The system according to this variant makes it possible to calculate a certain number of motor cursors representative of the activity of the monitored individual.

[0047] In particular, SMA sliders make it possible to characterize the average activity of the monitored individual. The calculation of these sliders over predetermined periods reflects the individual's dynamics and their average activity over a day, for example. These SMA sliders are measures of the sedentary behavior of the monitored individual and are calculated based on accelerations (measurements provided by the accelerometer), the magnetic field (measurements provided by the magnetometer) and angular velocities (measurements provided by the gyroscope).

[0048] The HA sliders allow the energy of the monitored individual to be characterized. The calculation of these sliders over predetermined periods reflects the individual's walking speed. These HA sliders are calculated based on accelerations (measurements provided by the accelerometer), the magnetic field (measurements provided by the magnetometer), and angular velocities (measurements provided by the gyroscope).

[0049] The HM cursors allow for the characterization of the harmony in the activities of the monitored individual. The calculation of these cursors over predetermined periods reflects the stability and symmetry of the individual's gait. These HM cursors are calculated based on accelerations (measurements provided by the accelerometer), the magnetic field (measurements provided by the magnetometer), and angular velocities (measurements provided by the gyroscope).

[0050] The HC cursors allow for the characterization of irregularities in the frequency domain during the activities of the monitored individual. The calculation of these cursors over predetermined periods reflects the stability and balance of the monitored individual's gait. These HC cursors are calculated based on accelerations (measurements provided by the accelerometer), the magnetic field (measurements provided by the magnetometer), and angular velocities (measurements provided by the gyroscope).

[0051] Advantageously and according to the invention, said module for determining said fall risk score includes an automatic calculation model trained to determine a fall risk score, this calculation model, said second model, having been trained by means of a training database, said risk score bank, which includes values ​​of said motor sliders and profile scores of a plurality of individuals associated with detected fall occurrences of said plurality of individuals.

[0052] According to this advantageous variant, the determination of an individual's fall risk score relies on an automatic calculation model trained using a learning database (called a risk score bank). This learning database consists of the motor cursors and profile scores of a plurality of individuals likely to fall, associated with detected fall occurrences for this plurality of individuals. This learning database is preferably the profile bank enriched with motor cursors.

[0053] To establish such a training set, a plurality of individuals were equipped with a unit of the system according to the invention to regularly calculate the motor cursors. These individuals were then observed (by a supervisor or using a dedicated device) to determine the occurrences of falls. The training set thus made it possible to associate, with each occurrence of a fall, the values ​​of the motor cursors calculated before the fall occurred, and the profile score of the individual. Once this first learning base is established, the system according to this variant can be used to detect the occurrence of a fall and thus improve and enrich the learning base.

[0054] The use of such an artificial intelligence model makes it possible to automatically detect patterns characteristic of a risk of falling.

[0055] The automatic calculation model implemented by the module for determining an individual's fall risk score can be of any type. It can be a supervised learning neural network, a support vector machine (SVM) or any other machine learning algorithm.

[0056] This module therefore makes it possible to assign to each individual equipped with the system according to the invention a fall risk score which depends on the value of the calculated motor cursors and on his own characteristics.

[0057] Advantageously and according to the invention, said fall risk calculation module includes an automatic calculation model trained to determine a fall risk, this calculation model, said third model, having been trained by means of a training database, said risk bank, which includes data representative of variations in fall risk scores of a plurality of individuals associated with occurrences of falls detected of said plurality of individuals.

[0058] According to this advantageous embodiment, the determination of an individual's fall risk relies on an automatic calculation model trained using a learning database (called a risk bank). This learning database consists of the motor parameters and fall risk scores of a plurality of individuals likely to fall, associated with detected fall occurrences for this plurality of individuals. This learning database is preferably the profile database enriched with the motor parameters and fall risk scores.

[0059] The use of such an artificial intelligence model makes it possible to automatically detect patterns characteristic of a risk of falling.

[0060] The automatic calculation model implemented by the module for determining an individual's fall risk score can be of any type. It can be a supervised learning neural network, a support vector machine (SVM) or any other machine learning algorithm.

[0061] This module therefore makes it possible to assign to each individual equipped with the system according to the invention a risk of falling which depends on the value of the calculated motor sliders, his fall risk score and his own characteristics.

[0062] The various automatic calculation models implemented by these advantageous variants of the invention can be enriched by the results provided by the system and by the users themselves by validating or rejecting the fall risk alerts triggered by the system.

[0063] Advantageously and according to the invention, the system further comprises a radio module configured to be able to transmit the data determined and calculated by said processing unit to a remote server.

[0064] According to this aspect of the invention, the processing unit can transmit the processing results to a remote server, and in particular the profile score, the fall risk score, the motor stabilizers, and the fall risk calculated by the various modules of the processing unit. The radio module also allows the system to be configured during its first use and for the system's presence to be verified. The remote server can be used to refine the processing of the processing unit or to perform some of the processing. In this respect, the processing unit can be partially or fully integrated into the equipment worn by the monitored individual or be trained partially or fully by a remote server.In the case where the processing unit is wholly or partially formed by a remote server, the radio module forms wireless communication means configured to transmit data from the sensors to said processing unit.

[0065] Advantageously and according to the invention, the system further comprises an audio module including a microphone and a speaker configured to allow an exchange of voice information between the individual and a remote operator.

[0066] According to this aspect of the invention, the audio module allows an audio alert to be transmitted to the individual, for example, when a risk of falling has been detected by the system. This radio module can also receive voice information from the user, for example, to deny a risk of falling or to confirm that the alert has been received. Other uses can be envisaged depending on the intended applications.

[0067] Advantageously and according to the invention, the system further comprises a human-machine interface configured so that said individual can interact with said system and / or a remote operator and receive fall risk notifications.

[0068] According to this aspect of the invention, the human-machine interface allows messages to be displayed to the monitored individual (visual notifications). This human-machine interface may also include emergency buttons that the monitored person can activate when they fall or need help.

[0069] The invention also relates to a method for predicting the fall of an individual comprising: - the acquisition of representative measurements of the individual's posture and / or movement from at least one accelerometer, one magnetometer and one gyroscope, - a processing of acquired measurements.

[0070] The method according to the invention is characterized in that it further comprises: - the determination of a data point, called a profile score, which is a function of data representative of the individual's specific characteristics, - a calculation of data, called motor cursors, based on acquired measurements, and representative of the individual's activity over a predetermined time interval, - a calculation of a data point, called the fall risk score, based on said motor cursors and said profile score, - a determination of a fall risk, based on a variation of said fall risk score beyond a predetermined threshold over a predetermined time interval.

[0071] The advantages and technical effects of the system according to the invention apply mutatis mutandis to a process according to the invention.

[0072] The invention also relates to a fall detection system for an individual comprising: - a fall detection module configured to detect the fall of said individual based on at least one measurement from at least one sensor worn by said individual that exceeds a predetermined threshold, - a fall prediction system according to the invention configured to determine a fall risk score for said individual, - a module for modifying said predetermined threshold of said fall detection module according to said fall risk score provided by said fall prediction system.

[0073] A fall detection system according to the invention thus makes it possible to adjust the fall detection threshold based on the fall risk score provided by the fall prediction system according to the invention. In this way, it is possible to adapt the fall detection system to the individual. Therefore, the sensitivity of the fall detection system and its alert policy can evolve and adapt automatically to the monitored individual. Furthermore, such a system can predict a fall before it occurs through the integration of the fall prediction system according to the invention.

[0074] The invention also relates to a fall prediction system, a fall detection system and a method, characterized in combination by all or part of the characteristics mentioned above or below. List of figures

[0075] Other objects, features and advantages of the invention will become apparent from the following description, given by way of non-limiting example only, and which refers to the accompanying figures in which:

[0076] [Fig. 1] is a schematic view of a system according to one embodiment of the invention,

[0077] [Fig.2] is a more detailed schematic view of a housing of a system according to one embodiment of the invention,

[0078] [Fig.3] is a functional schematic view of a processing unit of a system according to an embodiment of the invention.

[0079] Detailed description of an embodiment of the invention

[0080] In the figures, scales and proportions are not strictly observed for illustrative and clarity purposes. Furthermore, identical, similar, or analogous elements are designated by the same reference numerals in all figures.

[0081] Fig. 1 illustrates a system according to an embodiment of the invention comprising a case 10 worn by an individual 8 to be monitored and a data processing unit 100 for data acquired by sensors housed in the case 10.

[0082] In [Fig. 1], the housing 10 is integrated into a watch worn by the monitored individual. According to other embodiments, the housing can be integrated into a necklace, a pair of glasses, or any equipment worn by the individual being monitored.

[0083] As illustrated in [Fig. 2], this housing 10 contains sensors for acquiring representative measurements of the individual's posture and movements. These sensors include at least one accelerometer 12, one magnetometer 14, and one gyroscope 16.

[0084] The accelerometer 12 is preferably a three-axis accelerometer in order to be able to to obtain direct acceleration measurements on the three axes of a right-handed trihedron (x, y, z). In another embodiment, three single-axis accelerometers oriented relative to each other according to the aforementioned trihedron can be used. It is also possible to use a single single-axis accelerometer, but the measurement accuracy will be lower. The three-axis accelerometer allows obtaining, at each sampling instant tak, the triplet (ptak, aïak), as well as its magnitude defined as ilA il u-^2+«l2+<€2'

[0085] The magnetometer 14 is also preferably a three-axis magnetometer so as to be able to directly obtain field measurements on the three axes of the right-handed trihedron (x, y, z). According to another embodiment, three single-axis magnetometers oriented with respect to each other according to the aforementioned trihedron can be used. The three-axis magnetometer makes it possible to obtain, at each sampling instant tmk, the triplet (m^k, as well as its modulus defined as

[0086] The gyroscope 16 is also preferably a three-axis gyroscope so as to be able to directly obtain angular velocity measurements on the three axes of the right-handed frame (x, y, z). According to another embodiment, three single-axis gyroscopes can be used. oriented relative to each other according to the aforementioned trihedron. The three-axis gyroscope makes it possible to obtain the triplet at each sampling instant. , pf ), as well as its modulus defined as 11 n „ lnX 2 , „v 2 . - 2. 1 || G || t ,-Jg^ + g^ +gl

[0087] It is also possible to replace the gyroscope and accelerometer with an inertial measurement unit

[0088] The sampling times Gà? lmb G* may be different or identical, without this calling into question the principle of the invention.

[0089] The acquisition of data by the accelerometer 12, the magnetometer 14 and the gyroscope 16 as well as other sensors, if applicable, is controlled by a control card housed in the casing 10. This control may for example consist of defining the acquisition frequency (the sampling instants tab lmb ^gk mentioned previously).

[0090] The housing 10 also includes, according to the embodiment of [Fig. 2], a radio module 17 which can be used to transmit the data determined and calculated by said processing unit to a remote server, in the case where the processing unit is embedded in the housing. When the processing unit is remote, the radio module 17 can be used to initialize the system, verify the presence of the housing, etc.

[0091] The housing 10 also includes, according to the embodiment of [Fig. 2], a human-machine interface 18 configured so that the individual can interact with the system and / or a remote operator and receive fall risk notifications. This interface 18 can be of any known type and is not described in detail.

[0092] The housing 10 also includes, according to the embodiment of [Fig.2], a microphone and a loudspeaker 19 which form an audio module configured to allow an exchange of voice information between the individual and a remote operator.

[0093] According to an embodiment of the invention, not shown in the figures, the housing 10 can also include a pressure sensor and a temperature sensor to derive information related to the altitude of the individual wearing the housing.

[0094] The system may also include, according to an embodiment not shown in the figures, a sphygmomanometer configured to measure the blood pressure of the individual to be monitored, and / or a cardiac tachometer configured to measure the heart rate of the individual to be monitored.

[0095] The processing unit 100 can be integrated into the housing 10 and worn directly by the individual monitored by the system, or it can be located remotely and housed on a remote server. If the processing unit 100 is located remotely, wireless communication means 50 connect the sensors housed in the housing 10 to the processing unit 100. These wireless communication means include, for example, a 3G, 4G, 5G, WIFI connectivity, etc.

[0096] This connectivity not only allows the transmission of data from the box 10 to the processing unit 100, which is for example made up of a software application present on a remote server or of a set of software applications and databases present on a set of remote servers (also referred to by the terminology of cloud), but also allows the transmission of alerts or information to the box.

[0097] The embodiment described in connection with the figures includes a processing unit 100 consisting of a server remote from the device. However, nothing prevents the processing unit from being partially or totally integrated into the device worn by the monitored individual in other embodiments.

[0098] Thus, and according to the embodiment shown in the figures, the processing unit 100 receives the data from the sensors of the housing 10 by means of wireless communication 50 linking the housing 10 and the processing unit 100.

[0099] The [Fig.3] is a functional diagram of the processing unit 100 implemented by a system according to the invention.

[0100] The processing unit 100 includes, for example, a computing device 102, which should be understood in a broad sense (computer, plurality of computers, virtual server on the internet, virtual server on the cloud, virtual server on a platform, virtual server on a local infrastructure, server networks, etc.). This computing device typically includes one or more processors 106, one or more memories 108 containing instructions for software routines used by the system, and a human-machine interface 104.The processing unit also includes a database 116 for saving the results of the processing, accessing information relating to previous processing, and user data (which is provided for example via the human-machine interface 104 of the computer device 102 or via the human-machine interface 18 embedded on the box 10 and wireless communication means 50).

[0101] The processing unit 100 comprises four main modules: a module for determining the profile score of the monitored person 110; a module for calculating the motor cursors of the monitored person 120; a module for calculating a fall risk score of the monitored person 130; and a module for determining a fall risk of the monitored person 140.

[0102] In the following, the term "module" refers to a software element, a subset of a software program that can be compiled separately, either for independent use or for assembly with other modules of a program, or a hardware element, or a combination of a hardware element and a software subprogram. Such a hardware element may include an integrated circuit specific to a particular application. An application-specific integrated circuit (ASIC) or a field-programmable gate array (FPGA) or a specialized microprocessor circuit (DSP) or any equivalent hardware or combination thereof. Generally speaking, a module is therefore an element (software and / or hardware) that enables a specific function.

[0103] The processing unit 100 also includes means for storing the trained computing models 117, 118, 119 implemented by the invention. These means for storing the trained computing models may be servers separate from the computer device 102 or be stored within the computer device 102.

[0104] The automatic computing models implemented by the invention can be of different types. They can be a supervised learning neural network, a support vector machine (SVM) or any other machine learning algorithm.

[0105] The various modules of the computer device 102, described below, use in particular the processors 106, the memories 108, the database 116 and the means of storing the trained computing models 117, 118, 119, in order to be executed.

[0106] The profile score determination module 110 implements, according to the embodiment shown in the figures, an automatic calculation model 117 trained to determine a profile score for the individual, based on the individual's own characteristics, which are for example saved in the database 116. These own characteristics include for example one or more pieces of information related to the individual's age, sex, prescribed medications, weight, height, vision, hearing, history of falls, etc.

[0107] The training set used to train the model consists of the characteristics of a plurality of individuals likely to fall associated with occurrences of falls detected for this plurality of individuals.

[0108] To create such a training dataset, a number of individuals, whose specific characteristics used by the system (age, sex, weight, etc.) were recorded, were observed (by a dedicated team of operators or by equipping the individuals with a fall detection device) to detect occurrences of falls. All the information was saved in a database, forming the training dataset for the calculation model 117. Training model 117 allows it to associate a profile score with combinations of individuals' specific characteristics. In other words, model 117 makes it possible to determine the intrinsic risk factors for falls for an individual.

[0109] The motor cursor calculation module 120 uses the processor 106 and software routines stored in the memories 108 to evaluate the values ​​of the monitored person's motor cursors at predetermined time intervals. The data provided by the sensors is sent to the processing unit 100 via wireless communication means 50. Of course, and as previously indicated, it is also possible to have the motor cursor calculation module 120 directly integrated into the housing 10, in which case the motor cursors are saved in memory embedded in the housing 10 and / or transmitted to the remote database 116. Those skilled in the art will readily understand that the location of the various processing operations implemented by the invention is irrelevant and different variations are possible without calling into question the principles of the invention.

[0110] According to one embodiment of the invention, module 120 calculates the following set of motor cursors: - the cursors representing the average activity of the individual over a predetermined time interval T, called SMA cursors, calculated from the measurements provided by the accelerometer 12, the gyroscope 14 and the magnetometer 16 housed in the casing 10 according to the following formulas: • 4 1 / r+T| vl J vl J -IJ \ WHERE represent the acceleration values ​​measured on three axes of a right-handed trihedron (x,y,z) of said accelerometer, • i / rz+^i Ït+T i d+r \ where SMA m = ~ J t | | dt + J t | nij | dt + J t | / nf | dt represent the magnetic values ​​measured on three axes of said right-handed trihedron (x, y, z) of the magnetometer, • < / ft+T rt+T çt+T \ Q* Qy p' SMAG = |^Jf |^|^ + Jr f' r -z represent the gyroscopic values ​​measured on three axes of said right-handed trihedron (x, y, z) of the gyroscope, The cursors represent the individual's instantaneous activity, calculated from measurements provided by each of the sensors housed in the device, according to the following formulas: [YES]

[0112] It M || ( ++ the sliders representing the energy of said individual, called HA sliders, calculated as the variation of instantaneous activity over said predetermined period T according to the following formulas: 3 where var represents the variation of the value t / HA? = vari || over the predetermined time interval T, — Var[ || Àf || j ° where var represents the variation of the value over the predetermined time interval T, HA? = It G || f) where var represents the change in value over the predetermined time interval T, the cursors representing the harmony of the individual's activity, called HM cursors, over said predetermined time interval according to the following formulas: The cursors representing irregularities in the frequency domain during the individual's activities over the predetermined time interval T, called HC cursors, are measured according to the following formulas:

[0113]

[0114]

[0115] These different cursors are, for example, saved in database 116 so that they can be used by other modules of the system. Module 130 for calculating a fall risk score for the monitored individual implements, according to the method of realization of the figures, an automatic calculation model 118 trained to calculate a fall risk score, based on the profile score determined by module 110 and the motor cursors calculated by module 120. The training set used to train model 118 consists of the motor sliders and profile scores of a plurality of individuals associated with detected occurrences of falls from this plurality of individuals.

[0116] To create such a training dataset, a plurality of individuals, whose motor parameters were calculated at predetermined time intervals and whose specific characteristics used by the system (age, sex, weight, etc.) were recorded, were observed (by a dedicated team of operators or by equipping the individuals with a fall detection device) to detect occurrences of falls. All the information was saved in a database, forming the training dataset for the calculation model 118. Training the model 118 allows it to associate a fall risk score with combinations of specific characteristics and the motor parameters of the individuals.In addition, it is possible to regularly enrich the learning base from the users of the system according to the invention (the motor cursors and the specific characteristics of the individuals for whom the system has detected a risk of falling or a fall, and confirmed by the individual, are retrieved).

[0117] Module 140 for determining a fall risk implements, according to the embodiment of the figures, an automatic calculation model 119 trained to determine a fall risk from the variations of the fall risk score calculated by module 130, beyond a predetermined threshold.

[0118] The training set used to train model 119 consists of fall risk scores and motor sliders of a plurality of individuals associated with detected fall occurrences of this plurality of individuals.

[0119] To create such a training dataset, a plurality of individuals, whose fall risk scores and motor sliders were calculated at predetermined time intervals, were observed (by a dedicated team of operators or by equipping the individuals with a fall detection device) to detect occurrences of falls. All the information was saved in a database, forming the training dataset for the calculation model 119. Training the model 119 enables it to associate a fall risk with abrupt changes in the fall risk score.

[0120] This allows the monitored individual to be alerted to the high probability of a fall within a short period, typically 1 to 5 hours. This alert thus helps to prevent the fall from occurring.

[0121] The alert can be transmitted to the individual via the human-machine interface 18 carried by the device and / or by a voice alert transmitted to the audio module 19 and / or to a person defined by the user of the system as a reference person (a parent, a doctor, etc.) who can then contact the monitored individual to alert him of a risk of falling and advise him to reduce his activity in the coming hours.

[0122] A system according to the invention thus allows, through the combination of a set of sensors, different software routines implementing calculation models at automatic, to predict a risk of falling, whereas previous solutions were limited to detecting a fall.

[0123] The invention also extends to a fall detection system equipped with a fall prediction system according to the invention. In particular, the fall risks provided by the system according to the invention can be used as a parameter of the fall detection system, for example by adapting the detection thresholds to the fall risk information provided by the system.

Claims

1. Demands A fall prediction system for an individual, comprising: - a casing (10) configured to be worn by an individual (8), said casing (10) comprising a plurality of sensors for acquiring representative measurements of the posture and / or movements of the individual, including at least one accelerometer (12), one magnetometer (16) and one gyroscope (14), - a processing unit (100) for the measurements provided by said plurality of sensors, characterized in that said processing unit (100) comprises at least: - a module (110) for determining a data point, called a profile score, a function of data representative of the characteristics specific to the individual (8), said module (110) comprising an automatic calculation model (117) trained to determine a profile score, this calculation model (117), called the first model, having been trained using a training database, called the profile database, which includes data representative of the characteristics specific to a plurality of individuals associated with detected occurrences of falls of said plurality of individuals, - a data calculation module (120), called motor cursors, based on measurements provided by said plurality of sensors, and representative of the individual's activity (8) over a predetermined time interval, - a calculation module (130) for a data point, called a fall risk score, as a function of said motor cursors and said profile score, said module (130) comprising an automatic calculation model (118) trained to determine a fall risk score, this calculation model (118), called the second model, having been trained using a training database, called the risk score database, which includes values ​​of said motor cursors and profile scores of a plurality of individuals associated with detected fall occurrences of said plurality of individuals, a module (140) for determining a fall risk, a function of a variation of said fall risk score beyond a predetermined threshold over a predetermined time interval, said module (140) comprising an automatic calculation model (119) trained to determine a fall risk, this calculation model (119), called the third model, having been trained using a training database, called the risk bank, which includes data representative of variations in the fall risk scores of a plurality of individuals associated with detected fall occurrences of said plurality of individuals.

2. System according to claim 1, characterized in that said data representing the characteristics specific to each individual include one or more pieces of information related to the age, sex, prescribed medications, weight, height, sight, hearing, history of falls, etc. of said individual.

3. A system according to any one of claims 1 to 2, characterized in that said motor cursors calculated by said calculation module (120) are chosen from the group comprising: - cursors representative of the average activity of the individual over a predetermined time interval T, said SMA cursors, calculated from the measurements provided by each of the sensors housed in said casing according to the following formulas: — ( / çt+T rt+1 \ SMAa-y l]f +J t | [ dt + J t af, a?, ttj represent the acceleration values ​​measured on three axes of a right-handed trihedron (x,y,z) of said accelerometer, - , / ff+7\ i \ where SMAm = ÿ J t | mf । dt + J t | | dt + jt | | Jr mf, m^, represent the magnetic values ​​measured on three axes of said right-handed trihedron (x, y, z) of the magnetometer,— I / d+r rf+r ct+1 \ where Lg^+h kkk^+kg*, g?' 8zt represent the gyroscopic values ​​measured on three axes of said right-handed trihedron (x, y, z) of the gyroscope, - cursors representing the instantaneous activity of, The individual is calculated from the measurements provided by each of the sensors housed in said casing according to the following formulas: He A / || It G II || A || t = y]af + af + af' sliders representing the energy of said individual, called HA sliders, calculated as the variation of instantaneous activity over said predetermined period T according to the following formulas: HAy — var( Il A II ) where var represents the variation of the value over the predetermined time interval T, HA^ - varÇ || M || ) where var represents the variation of the value over the predetermined time interval T, HA^~ var^ Il G II j °ù var represents the variation of the value over the predetermined time interval T, of the cursors representing the harmony of the individual's activity, called cursors HM, over said predetermined time interval according to the following formulas: II has 11, = ^4 hm^ H.!ri!!J ott II m || , = ^ sf ii g 11,=^ T Mm,) Sliders representing irregularities in the frequency domain during the individual's activities over the predetermined time interval T, called HC sliders, are measured according to the following formulas: „rc Hlû'l,) where |O'| HC,= i ---------r- 11 II t Af 7 YMiMÜ

4. System according to any one of claims 1 to 3, characterized in that it further comprises a radio module configured to be able to transmit the data determined and calculated by said processing unit to a remote server.

5. System according to any one of claims 1 to 4, characterized in that it further comprises an audio module (17) comprising a microphone and a loudspeaker configured to allow an exchange of voice information between the individual and a remote operator.

6. System according to any one of claims 1 to 5, characterized in that it further comprises a human-machine interface (18) configured so that said individual can interact with said system and / or a remote operator and receive fall risk notifications.

7. A method for predicting the fall of an individual comprising: - an acquisition of measurements representative of the posture and / or movement of the individual from at least one accelerometer, a magnetometer and a gyroscope, characterized in that it further comprises: - a determination of a data, called profile score, as a function of data representative of the individual's own characteristics, - a calculation of data, called motor cursors, as a function of the acquired measurements, and representative of the individual's activity over a predetermined time interval, - a calculation of a data, called fall risk score, as a function of said motor cursors and said profile score, - a determination of a fall risk, as a function of a variation of said fall risk score beyond a predetermined threshold over a predetermined time interval.

8. A fall detection system for an individual comprising: • a fall detection module configured to detect the fall of said individual based on at least one measurement from at least one sensor worn by said individual that exceeds a predetermined threshold, a fall prediction system according to any one of claims 1 to 6 configured to determine a fall risk score of said individual, a module for modifying said predetermined threshold of said fall detection module according to said fall risk score provided by said fall prediction system.

Images