EP3488379A1

EP3488379A1 - Method of processing data to improve motion recognition, associated sensor and system

Info

Publication number: EP3488379A1
Application number: EP17740738.4A
Authority: EP
Inventors: Fernando Romao; Andrei SHELEH; Ivan SELIVANAU
Original assignee: Octonion SA
Current assignee: Octonion SA
Priority date: 2016-07-22
Filing date: 2017-07-17
Publication date: 2019-05-29
Also published as: US20210287057A1; FR3054346A1; WO2018015355A1

Abstract

The invention relates to a system and a method for processing data so as to improve motion recognition implemented by an application associated with at least one data sensor (1; 1'). Raw data are collected (E₄) by said at least one sensor and stored (E₁₀) in a remote database (7). An analysis, preferably, statistical (E₁₂) of all or part of the raw data stored in the database (7) is performed by a server (5) linked to the database (7). The application is parametrized and/or updated (E₁₄) as a function of the results of said analysis (E₁₂). The system comprises one or more data sensors (1; 1 '), the server (5) and the remote database (7) and a mobile terminal (3).

Description

Data processing method for improving motion recognition, associated system and sensor.

Field of the invention

The present invention relates to the field of motion recognition implemented by data sensors.

The invention finds a privileged application in the field of sport to recognize the movements made by an athlete in different disciplines, such as golf, tennis, or skiing.

In this field of application, there are so-called data sensors

"Multisports" for evaluating the performance of an athlete during his sport, these sensors can be configured according to the sport practiced.

For example, during a tennis training session, a data sensor attached to an athlete's wrist should be used to identify how many strokes or backhands were made by that athlete. For this, a reliable recognition of the movements made by the athlete is necessary.

In view of the multitude of sports practiced around the world, there is a need to configure or adapt efficiently the motion recognition algorithms implemented on each sensor, to improve the reliability of recognition and thus make more reliable the resultant sports performance analysis.

According to the state of the art, it is known to locally exploit raw data, such as inertial quantities measured by accelerometers and / or gyrometers embedded in multisport data sensors. These data are processed directly by the data sensor by means of a dedicated application implementing motion recognition algorithms, these algorithms being designed to recognize the nature or type of movements performed during a sporting practice.

The dedicated application analyzes the raw data provided by the sensors and produces so-called reduced data which contain the recognized movement and possibly associated measurements. For example, the reduced data may contain the information that the recognized movement is a forehand to tennis associated with the speed of the racket at the moment of the typing. However, it may happen that the recognition of movements by the algorithms implemented by the dedicated application is not optimal. This may come from physical specifics or from the practice of the user for example.

Object of the invention

The present invention aims to remedy the aforementioned drawbacks, by proposing a technical solution to improve the recognition of movements implemented by data sensors, for a wide variety of sports practices.

These objects are achieved by a data processing method intended to improve the motion recognition implemented by an application associated with at least one data sensor, said method comprising the following steps:

collecting raw data from said at least one sensor;

storing raw data in a remote database; analyzing, in particular statistics, all or part of the raw data stored in the remote database; and

parameterizing and / or updating said application based on the analysis results obtained at the end of the analysis step.

The analysis, in particular statistics, raw data from a plurality of sensors used by different users is particularly advantageous for improving the motion recognition performance of a generic algorithm of the application, particularly when such an algorithm is implemented by a large number of sensors. The use of statistical analysis methods is all the more advantageous as the number of sensors for feeding the remote database is high.

For a given sensor, the analysis, in particular statistics, raw data from the same sensor is also advantageous for improving the performance of a motion recognition algorithm of the application specifically to this sensor, the algorithm can be Customized based on the raw data provided by this sensor.

In any case, it is on the basis of the raw data obtained from one or more data sensors that the recognition of movements of an athlete can to be improved. The result of the analysis is used to parameterize or update the motion recognition algorithm (s) centrally. This is all the more advantageous as the number of deployed sensors is high. Thus, the motion recognition applications implemented on the data sensors can be updated or configured centrally, from the raw data obtained from a set of sensors.

According to a particular embodiment of the invention, the method further comprises a preliminary test step, during which a criterion of reliability of motion recognition is evaluated, since a movement is recognized by the application, of so that the other steps of the method are initiated according to the value of the reliability criterion with respect to a predetermined threshold.

Thus, the steps of the method according to the invention are executed only in the case where the motion recognition performed by the data sensor has an insufficient degree of reliability. The conditional execution of the steps of the method according to the invention is particularly advantageous to avoid unnecessary processing and preserve the electrical resources of the data sensor whose autonomy must be optimized.

Since the raw data is usually large, the fact that it is sent on an ad hoc basis is particularly advantageous for limiting the amount of information to be transmitted through the radio communication networks, given the limited transmission capacities of the data carriers. radio technologies currently deployed and massive deployment of connected data sensors.

According to a particular embodiment of the invention, an identifier qualifying the type of movement performed by a user of said at least one sensor is associated with said raw data during the collection step. The identifier is entered, manually by the user or automatically, according to a sequence of movements to be made predefined, the identifier and the data being transmitted to said database.

The identifier advantageously makes it possible to classify and group the raw data of the same nature (for example, corresponding to the practice of the same sport under similar conditions) in the remote database. According to a particular embodiment of the invention, the collected raw data are sent by said at least one data sensor to a mobile terminal, this terminal being adapted to transmit said data to the remote database.

The use of a mobile terminal, such as a mobile phone, to transfer raw data from the data sensor to the remote database is particularly well suited in the case where the data sensors do not have means to communicate. with mobile radio networks.

Thus, the raw data collected is transmitted to the remote database by the mobile phone using the subscription linked to this phone. In this case, it is possible to use simplified architecture sensors, whose energy consumption is reduced.

In an exemplary embodiment, the raw data comprises acceleration and / or speed inertial data.

The inertial data is not only used by the motion recognition algorithm but also exploited during the analysis step to adapt the application of the data sensor, in order to improve the recognition capabilities.

Optionally, the raw data further comprises at least one of the following parameters: data sensor orientation, ambient pressure, ambient temperature.

These additional parameters are indicators of the physical conditions in which the data sensor was used during the collection of inertial data. Such indicators may be taken into consideration when analyzing the raw data collected.

According to a particular embodiment, the method further comprises a merging step, during which the raw data collected is merged, so as to provide a calibrated, filtered and more accurate data structure than if considered in isolation.

As a first example, the gyro biases (ie low frequency component) are suppressed by calculating a sliding average in the static phases of the data sensor or are removed by more powerful algorithms such as those described by Kalman-Bucy. and known as "Kalman Filters". As a second example, the data of the gyrometer and the accelerometer are combined, that is to say, merged to achieve greater precision on the calculation of the attitude (orientation in space) of the data sensor. . This merger can be done by Kalman filtering or by using a complementary filter. These techniques are known to those skilled in the art and therefore are not described in detail here.

Thus, the function of the merging of the collected raw data is to calibrate the quantities measured during data collection, for example, with respect to a common reference value and / or to convert these data so as to express them according to a system of measurement. conventional units directly interpretable during the analysis step. The calibration makes it possible to compensate for static and dynamic defects by combining complementary data to improve the overall accuracy.

According to a particular embodiment, the parameterization step includes supplying said at least one data sensor with at least one configuration parameter determined according to the results of the analysis.

The configuration parameter (s) provided to the data sensor (s) are used to customize the recognition algorithm implemented by the application of this data sensor (s). The centralized provision of these configuration parameters is particularly advantageous for simultaneously customizing a plurality of deployed sensors, without having to intervene individually on each of the sensors.

According to a particular embodiment, the update step includes providing at least one sensor of an updated application based on the results of the analysis.

Thus, a centralized update of the applications of the sensors can be carried out efficiently. This is particularly advantageous for generalizing a recognition algorithm (i.e. improving a generic algorithm) on the basis of raw, preferably merged, data obtained from a large number of data sensors.

The invention also relates to a data processing method intended to improve the recognition of movements implemented by an application associated with at least one data sensor. Said method comprises the following steps implemented by said at least one data sensor:

raw data collection;

sending the collected raw data to a remote database; and - receiving configuration parameters and / or updating said application, said parameters and / or said updating being determined according to the results of an analysis processing, preferably statistical, of all or part raw data stored in the remote database. The invention also relates to a data processing method intended to improve the recognition of movements implemented by an application associated with at least one data sensor, said method comprising the following steps implemented by a data server:

receiving raw data collected by said at least one sensor;

storing said raw data in a remote database;

- analysis, preferably statistical, of all or part of the data contained in said database; and

parameterizing and / or updating said application according to the results of the analysis step.

The invention also aims at a data processing system intended to improve the recognition of movements implemented by an application associated with at least one data sensor, said system comprising:

means for collecting raw data by said at least one sensor; means for storing the raw data in a remote database;

means for analyzing, preferably statistically, all or part of the raw data stored in the database; and

means for parameterizing and / or updating said application as a function of analysis parameters obtained by the analysis means.

According to a particular embodiment of the invention, the system further comprises a mobile terminal adapted to transmit to the remote database collected raw data received from said at least one sensor. The invention also relates to a data sensor for improving the recognition of movements, said sensor comprising:

means for collecting raw data;

means for sending said data to a remote database;

means for receiving at least one configuration parameter and / or updating said application, said at least one parameter and / or said update being determined according to the results of an analysis, preferably statistic, all or part of the raw data stored in the remote database.

The invention also aims at a data server intended to improve the recognition of movements implemented by an application associated with at least one data sensor, said server comprising:

means for receiving raw data obtained by means for measuring at least one data sensor;

means for storing said raw data in a database; means for analyzing, preferably statistically, all or part of the data contained in said database; and

means for parameterizing and / or updating said application as a function of the results obtained by the analysis means.

According to a particular embodiment, the server is adapted to implement at least one motion recognition algorithm according to an algorithm implemented on said at least one data sensor to be parameterized or updated. Thus, the algorithm is executed by the server using all or part of the raw data stored in the database.

The invention also relates to a computer program comprising instructions adapted to the implementation of any of the steps of the methods according to the invention as described above, when said program is executed on a computer.

This program can use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in a partially compiled form, or in any other form desirable shape. The invention also relates to an information storage medium, removable or not, partially or completely readable by a computer or a microprocessor comprising code instructions of a computer program for the execution of any one of the steps methods according to the invention as described above.

The information carrier may be any entity or device capable of storing the program. For example, the medium may comprise storage means, such as a ROM (Read Only Memory), for example a microcircuit ROM, or a magnetic recording means, for example a hard disk, or a memory flash.

On the other hand, the information medium may be a transmissible medium, such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by other means. The program according to the invention may in particular be downloaded to a storage platform of an Internet type network.

Alternatively, the information medium may be an integrated circuit, in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of the method in question.

The aforementioned information carrier and computer program have characteristics and advantages similar to the method they implement. Brief description of drawings

Other features and advantages of the invention will become apparent in the description below, in relation to the accompanying drawings, given by way of non-limiting examples, and in which:

Figure 1 illustrates the architecture of a system according to a first embodiment of the invention;

Figure 2 illustrates the architecture of a system according to a second embodiment of the invention;

Figure 3 illustrates an exemplary embodiment of the system according to the first embodiment of the invention comprising a plurality of sensors;

Figure 4 illustrates the steps of the method according to a particular embodiment of the invention; Figure 5a illustrates an example of representation of two point clouds corresponding respectively to two gestures by illustrating the Hamming distance; and Figure 5b illustrates an example of partitioning to allow motion recognition.

One of the ideas underlying the invention is to parameterize and / or update the application of one or more data sensors to improve the reliability of motion recognition, depending on the results of an analysis, preferably statistical, raw data stored in a remote database, these data being collected from at least one data sensor.

The system according to the invention comprises at least one data sensor used by one or more users, a server connected to the remote database. The server is adapted to analyze the raw data contained in the database. The collected data can be sent to the server via a mobile terminal connected to the data sensor. In an alternative embodiment, the data sensor is adapted to transmit the collected data directly to the data server.

The raw data stored in the database is associated with an identifier for qualifying the type of movement made during the collection of corresponding raw data. Each set of raw data collected in association with the predefined motion identifier constitutes a complete block of data, all the blocks of data stored in the database that can be exploited by statistical preference analysis methods.

The results of the raw data analysis are used to determine configuration parameters for improving the configuration of the motion recognition algorithm implemented by the data sensor application. Thus, the invention advantageously allows a reconfiguration of the application of each data sensor, taking into account all or part of the raw data stored in the remote database.

The results of the raw data analysis can also be used to develop an update of the application of the data sensor, on which the motion recognition algorithm is implemented. Thus, the invention makes it possible to update the application of one or more data sensors, taking into account the all or part of the raw data stored in the remote database obtained from the sensors.

In any case, the configuration and / or update of the application of one or more data sensors are performed centrally, by means of the server, taking into account all or part of the raw data obtained from a sensor or a plurality of sensors.

Typically, the invention makes it possible to update and / or configure a generic motion recognition algorithm, this algorithm being able to be common to a plurality of sensors regularly supplying the remote database.

The purpose of the algorithms deployed on all the sensors is not only to recognize a given movement and categorize it (for example, forehand tennis) in order to create statistics communicated to the user, but also to measure with the greatest possible accuracy a characteristic parameter that will also be communicated to the user, such as the speed of a movement or the height of a jump.

Figure 1 schematically illustrates the architecture of a system according to a first embodiment of the invention.

The system includes:

a multisport data sensor 1,

a mobile terminal 3 constituted, for example, by a mobile phone, - a server 5 and a remote database 7 in a telecommunications network 9.

In the present example, the telecommunications network 9 comprises the Internet network in which the server 5 and the remote database 7 appear. The mobile terminal 3 is equipped with 3 G or 4 G radio-mobile communications means adapted to access to the Internet according to IP communication protocols.

The data sensor 1 comprises measurement means Mi, ..., M _n for measuring inertial variables, such as speed and / or acceleration, these measuring means being integrated in the data sensor 1.

In a particular embodiment of the invention, the data sensor 1 comprises three measuring means as follows: first measuring means Mi adapted to measure an instantaneous angular velocity, this velocity being represented by a vector ω having three rotational speed components co _x , co _y , co _z around each of the axes of said first measuring means Mi constituted by a gyrometer embedded in said data sensor;

second measuring means M ₂ adapted to measure a local linear acceleration represented by a vector γ having three linear acceleration components γ _χ , y _y , γ _ζ along each of the axes of said second measuring means M ₂ , these means being constituted by a linear accelerometer embedded in said data sensor;

third measurement means M3 adapted to calculate a universal acceleration Γ (Linear World Acceleration) from the accelerations measured by one or more accelerometers and rotational speeds measured by one or more gyrometers, the universal acceleration being calculated according to known algorithms those skilled in the art (eg rotation matrices using Euler angles or quaternions, data fusion using complementary filters, Kalman filtering), these means creating what can be called "universal accelerometer".

The data sensor 1 may further comprise additional measuring means, such as a pressure sensor and / or a temperature sensor.

The data sensor 1 further comprises:

a central processing unit 12 comprising a microprocessor, in particular of ARM type such as Cortex-M4 or other;

a random access memory 13 of the Random Access Memory (RAM) type and a Read Only Memory (ROM) type ROM 14;

a communication interface 15, for example of the short-range radio frequency type of Bluetooth® type.

All the constituent elements of the data sensor 1 as described above can be typically integrated on the same printed circuit board PCB type monobloc.

In the example described here, the read-only memory 14 constitutes a recording medium which stores a computer program PG1 conforming to the invention, able to implement, when executed by the central unit 12, the steps of the data processing method performed by the data sensor in accordance with the invention, as illustrated in FIG. 4.

The mobile phone 3 typically includes:

a Bluetooth® type short-range radio communication interface (COM1) adapted to communicate with the communication interface 15 (COM) of the data sensor 1;

a type 3 G or 4G radio-mobile communication interface 36 (COM2);

RAM RAM 33 and ROM ROM 34;

a central processing unit 32, such as a microprocessor;

a human-machine interface 37 (HMI) comprising a display device.

In the example described here, the ROM 34 (ROM) constitutes a recording medium which stores a computer program PG3 according to the invention, able to implement, when executed by the central unit. 32, the steps of the data processing method performed by the mobile terminal 3 according to the invention, as illustrated in FIG. 4.

The server 5 is connected to the database 7 on the telecommunications network 9. The server 5 typically comprises:

a communication interface 55 (COM) adapted to communicate with the mobile radio communication interface 36 (COM2) of the mobile phone 3;

a random access memory 53 of the RAM type and a read-only memory 54 of the ROM type;

a central processing unit 52, such as a microprocessor.

In the example described here, the read-only memory 54 constitutes a recording medium which stores a computer program PG5 according to the invention, able to implement, when executed by the central unit 52, the steps of the data processing method performed by the server 5 according to the invention, as illustrated in FIG. 4.

Figure 2 schematically illustrates the architecture of a system according to a second embodiment of the invention. This second embodiment differs essentially from the first embodiment described with reference to FIG. 1, in that the data sensor 1 'is adapted to communicate directly with the server 5. In this case, it is not necessary to have a mobile phone 3, which is the case for autonomous multifunction data sensors.

In this case, the data sensor 1 'comprises, for example, a type 3G, 4G and / or wireless type Wi-Fi® communication interface 15' (COM) adapted to communicate directly with the server. 5.

The data sensor further comprises a human-machine interface 17 (IMH) comprising a display device.

In order not to unnecessarily overburden this description, all the other elements designated by the same reference numbers as those already described with reference to FIG. 1 remain identical and are not described again.

FIG. 3 illustrates an embodiment of the system according to the first embodiment of the invention, in which a plurality of data sensors 1.1, 1.2a, 1.2b,..., 1n are adapted to feed the same remote database 7 via mobile phones 3.1, 3.2, ..., 3.n with which they are respectively associated. It will be noted that several sensors 1.2a, 1.2b can be associated with the same mobile phone 3.2.

FIG. 4 illustrates an exemplary embodiment of the method according to the invention as implemented by the system illustrated in FIG. 1 according to the first embodiment.

For the sake of simplification, the method will be described with reference to a single data sensor, but will remain valid for each of the plurality of sensors of the system shown in FIG. 3, where multiple sensors may simultaneously or sequentially transmit raw data to the remote database 7. The server 5 and the database 7 are sized appropriately according to the number of sensors and / or mobile phones that can feed the database 7.

Initially, it is assumed that the data sensor 1 is associated with the mobile phone 3 by means of the application PG3 executed on the telephone 3, during a prior association step Eo (pairing mechanism in English). As soon as the data sensor 1 is associated with the telephone 3, the user can select from the application PG3 the type of sport practiced by means of the human-machine interface 37. This information is entered in the sensor 1 which can be configure according to the selected sport.

In a nominal operating mode, the data sensor 1 executes the application PG1 intended to recognize, in real time, movements made by an athlete in action. For this purpose, the processor 12 of the data sensor 1 executes a predefined motion recognition algorithm in the application PG1, during a recognition step Ei.

This recognition is performed in real time and as a function of the raw data measured by the measurement means of the data sensor 1. In the present example, these raw data comprise the angular velocity values ω measured by the gyrometer Mi, the values of local linear acceleration γ measured by the local linear accelerometer M ₂ and the universal linear acceleration values Γ provided by the universal linear accelerometer M ₃ .

To enable motion recognition, the application PG1 implements a motion recognition algorithm, this algorithm being predefined during commissioning of the sensor using standard parameters.

However, according to a feature of the invention, this algorithm can be adapted or updated based on a statistical analysis of the raw data provided by one or more sensors as described below.

After each iteration of the recognition step Ei, the application PG1 verifies the reliability of this recognition, according to at least one reliability criterion, during a first optional test step E ₂ . For this purpose, the application PG1 compares the value of said at least one recognition reliability criterion with respect to a predetermined reference threshold (reliability condition).

As an illustrative and non-limiting example, the reliability criterion or indicator used is the Hamming distance. In this case, the first test step E ₂ consists of calculating the Hamming distance between the motion pattern recognized at a given instant and two neighboring patterns previously recognized at the end of the recognition step. If the calculated distance is greater than a predetermined reference threshold, the reliability condition is fulfilled and in this case, the data sensor continues the recognition Ei in real time of the movements made by the sportsman, in his nominal mode of operation.

If the computed Hamming distance is abnormally low, that is to say if it is too frequently below the predetermined reference threshold, a collection step E ₄ of the raw data is automatically initiated by the data sensor 1.

For example, the raw data collection step will be triggered provided that the Hamming distance is less than the reference threshold for more than 70% of the movements detected during a training session.

In this example, the rate of occurrence of this condition is set to 70%. This value can be defined on the basis of previously collected data and analysis of the statistical distribution of the population studied (used to build the motion recognition algorithms).

The recognition information obtained at the end of the recognition step Ei are reduced data {e.g. reverses, speed, effect) that can be transmitted to the application PG3 executed by the mobile phone 3 to indicate to the user the recognized movements and their characteristics, by means of the human-machine interface 37. Unlike the raw data, the reduced information can not be exploited to improve the motion recognition algorithms because the reduced data is the result of this recognition.

Optionally, the collection step E ₄ is initiated according to the profile of the user (sportsman) and / or his sports performance, after a second test step E ₃ .

For example, the collection step E ₄ can be initiated if at least one of the following conditions is met:

the user belongs to a predetermined category of athletes (e.g. professional);

the Hamming distance is too often below a threshold.

Such conditions are verified by the application PG1 of the data sensor 1 during the second test step E ₃ , as a function of profile information and / or performance of the user, all or part of this information that can be provided by the PG3 application executed on the mobile phone 3.

Depending on the type of sport previously selected by the user, the application PG3 presents to it by means of the human-machine interface 37 a predetermined sequence of movements to be performed during the data collection step E ₄ .

For example, in the case of tennis, the PG3 application will provide the user with the instructions to perform the following sequence of movements: five backs, five forehands and three services. Thus, the collected data can be associated with predetermined movements according to the proposed sequence.

This sequence comprises a list L of identifiers qualifying the type of movements to be performed. For example, this list L is pre-registered in the application PG1 of the sensor 1. It can also be transmitted to the sensor by the application PG3 of the mobile phone 3 when a sport has been selected on the PG3 application. Thus, an identifier I specifying the type of movement to be performed is automatically associated with the raw data collected according to the motion sequence presented to the user.

According to the example illustrated in Figure 3, it is assumed that at a given moment, a first identifier Ii denoting, for example, a tennis lapel is associated with the raw data Dl. l and D 1.2a obtained from the data sensors 1.1 and 1.2a following the execution of a setback by each of the two sportsmen. This data is transmitted to the server 5 in association with the first identifier Ii by mobile phones 3.1 and 3.2 respectively. Similarly, a second identifier I ₂ , denoting, for example, a tennis forehand is associated with the raw data D1.2b and Dl .n provided by the data sensors 1.2b and l .n, these data being transmitted to the server 5 by mobile phones 3.2 and 3.n respectively.

Thus, the data Dl. l and D1.2a associated with the first identifier Ii and the data D1.2b and Dl .n associated with the second identifier I2 are stored in the database 7.

According to an alternative embodiment, the raw data collected is manually qualified by the user, for example, when the latter selects in the application PG3 a movement from a predefined list of movements before to perform this movement during the collection step E ₄ . This qualification has the effect of associating the identifier I of the movement made with the raw data collected during the realization of this movement.

In the second embodiment, in the absence of the use of a mobile phone 3, the sequence of movements is presented by means of the human-machine interface 17 included in the data sensor.

In all cases, the nature of the movement realized by the athlete is qualified and associated with the raw data collected during the realization of this movement.

During the collection step E ₄ , these data are recorded by the data sensor, for example in the form of a local file. In the present example, the local file contains all the data measured by the gyrometer Mi, the local linear accelerometer M ₂ , the universal linear accelerometer M ₃ and possibly pressure and temperature data obtained by pressure sensors. and temperature during the collection step E ₄ .

Collected raw data is packaged during a packaging step

Isa. The packaging of the data consists in converting the measured data into a raw data structure calibrated with respect to reference values and expressed according to units that can be interpreted for example by a human or by a statistical processing algorithm. In this example, the packaged data constitutes a raw data structure comprising:

angular velocity data Di measured by the gyrometer Mi and expressed along three orthogonal axes in a coordinate system linked to the data sensor 1;

local acceleration data D ₂ measured by the local linear accelerometer M ₂ and taking into account the gravity and expressed in m / s ² along the three axes of the accelerometer in the local coordinate system linked to the data sensor 1 ; and

universal acceleration data D ₃ measured by the universal linear accelerometer M ₃ , not taking into account the gravity and expressed in m / s ² along the three axes of the accelerometer in the universal coordinate system.

The packaged data may also include: a special data structure called "Quaternion" describing the orientation of the data sensor in the space; and or

atmospheric pressure data expressed in Pascal (Pa); and / or atmospheric temperature data expressed in degrees Celsius (° C). This packaged data constitutes a data structure comprising inertial data and any raw pressure and temperature data, all of which data are calibrated with respect to reference values for each of the quantities concerned.

The raw data collected and packaged D are sent by the data sensor 1 to the mobile telephone 3 by means of a Bluetooth® communication, during a first transmission step E ₇ .

For each movement carried out during the collection step E ₄ , the packaged raw information D is sent by the communication interface 15, these data being sent in association with the identifier I making it possible to identify the nature or the type of the realized movement. According to the example of FIG. 3, the data bundled With associated with the identifier Ii provided by the data sensor 1.1 are transmitted by the mobile phone 3.1. to the data server 5.

On reception, the data Du, Di.2a, Di. _n are stored on the mobile phone 3 before being transferred by it to the server 5 via the network 9, during a second transmission step Es.

In the present exemplary embodiment, the sending of the raw data is automatically proposed to the user on the application PG3 executed on his mobile phone 3. In this case, the application PG1 executed on the data sensor 1 sends a request to the PG3 application of the mobile phone to ask the user to accept sending the collected raw data to the remote database 7.

According to an alternative embodiment, the sending of the raw data is performed without user intervention, for example randomly or automatically when the condition of reliability of recognition is not met.

Randomness is particularly advantageous for increasing the knowledge of the user base and refining the algorithms accordingly. For example, assuming that in a particular world region or for a category number of people given the recognition performance of the movements are less good, the obtaining of additional raw data is particularly advantageous to achieve a campaign creation and / or modification of motion recognition algorithms. The server can contact a dozen users of the category mentioned for the campaign via the PG3 application running on their mobile phone. In this case, a request is sent by the server to these phones. In response to this request, the user informs by means of the PG3 application the identifier I to identify the nature or type of movement made, so that it is transmitted in association with the raw data.

According to another variant embodiment, it is envisaged that the user wishing to improve the recognition of movements carried out by his data sensor can initiate of his own volition the steps of collection E ₄ , packaging Es and sending E ₇ raw data, from the PG3 application executed on its mobile phone 3, without the implementation of the optional test step E ₂ . In this case, the user informs the nature or type of movement made, so that it is transmitted in association with the raw data.

In the second embodiment, in the absence of a mobile phone, the collection of raw data is initiated by the user by means of an interface provided on the data sensor, which data is then sent by the data sensor 1 directly to the remote database 7. In this case, the sensor application includes all or part of the features provided on the PG3 application of the mobile phone described above.

In any case, the raw data packaged is not sent continuously, which limits the amount of data transmitted. This is particularly advantageous since the raw data is bulky.

For example, using an accelerometer clocked at a sampling rate of 500 Hz, the characterization of a motion for a duration of one second requires several kilobytes of data.

A one-off sending of these data significantly reduces the power consumption of the data sensor compared to the case where the data raw would be sent continuously. Thus the energy autonomy of the sensor can be optimized.

In addition, given the proliferation of connected data sensors, it is important to ensure that they do not unnecessarily encumber radio-mobile communication networks with limited transmission capacity. In this respect, sending the raw data under conditions is advantageous.

On receipt of the raw data, the server 5 stores this data in the remote database 7 during a storage step Eio. Since these raw data are associated with known movements according to the sequence of movements presented to the user during the collection step E ₄ , these constitute a set of data that can be used to increase the reliability of the motion recognition. implemented by different data sensors.

The remote database 7 comprises for each raw data from a sensor an identifier of the type of movement concerned. For example, the database may include several sets of characteristic data of the same type of movement for a given sport. According to the example illustrated in FIG. 3, the raw data D1.2b and D1 .n issued respectively from the sensors 1.2b and 1n are stored in the database 7 in association with the same identifier I ₂ designating, for example , a tennis lapel.

The centralized storage in the database 7 allows all the deployed sensors to contribute to the enrichment of the content of this database during the collection step E ₄ . Feeding this database centrally by a large number of sensors makes it possible to constitute a data set sufficiently complete to be analyzed statistically.

The enrichment of the content of the database makes it possible in particular to constitute an extensive list of possible ways to perform different movements.

The regular enrichment of the content of this database by raw data from different sensors is particularly advantageous for improving the overall reliability of motion recognition implemented on one or more sensors.

The server is adapted to implement at least one motion recognition algorithm for each sport practiced. For this, the server is adapted to implement at least the following three modules: an activity qualifier (activity qualifier) for identifying the nature of a movement characteristic of a sporting practice, by detecting the different phases of a movement from the measured raw data, a corresponding activity data stream being provided as output of the activity qualifier;

- a classifier (classifier) to classify the activity data from the qualifier; and

an analyzer for processing classified activity data from the classifier.

The activity qualifier provides the classifier with a flow of activity data including qualification values called qualifier values, such as rms and / or average values of acceleration and rotational speed along one or more axes, expressed for example in the local coordinate system at the sensor and / or in the coordinate system of the terrestrial reference.

The choice of qualifier values is made according to each sport considered, for example, following a human analysis. This choice can be made according to tests / errors type tests supported by statistical correlation analyzes.

For example, in the case of a movement made by a boxer, the qualifier detects, from the measured raw data, that a punch has been made by the boxer and provides the classifier acceleration data according to the local X axis (component γχ) measured by the local linear accelerometer M ₂ .

Depending on the sport considered, other examples of qualifying values may be considered, such as the amplitude of the movement and / or the projection elements of the trajectory of the movement in a reference plane.

For example, the trajectory detected by the sensor may be projected in a horizontal plane and the two dimensions (length L1 and width L2) of a rectangle in which the projected trajectory is written will be calculated. In this case, the length L1 and the width L2 of the rectangle are considered qualifier values which prove to be very useful to make the difference between a direct punch and a boxing hook, for example.

Other qualifying values may also be obtained by proceeding in the same way, following a plane perpendicular to the horizontal plane. In this case, the main plan motion is detected then the motion path is projected on this main plane.

According to an alternative embodiment, the trajectory may be temporally divided into a plurality of segments, so that qualifier values as described above are calculated on each of these segments.

To ensure good reliability of movement recognition, the classifier can be adapted to the sport practiced. This module allows the recognition of forms (pattern recognition) by the implementation of algorithms of artificial learning (machine learning), from qualifier values provided by the qualifier of activities. Such algorithms are based, for example, on C45 decision trees, logistic regression or neural networks. This classification method is supervised, in the sense that the nature of the movement is known beforehand to carry out the classification thanks to the flow of activity data provided by the classifier.

A linear regression algorithm can also be used to classify the different movements. Classification rules can be established using decision tree learning methods. The decision tree thus determined makes it possible to classify the movements, that is to say, to recognize them by relying on a large database of already classified sporting gestures.

The result provided by the classifier is a set of recognized motions which are generally associated with the acceleration and attitude data of the sensor in the terrestrial frame. These data are used by the analyzer implementing algorithms specific to each sport and each movement recognized by the classifier to calculate metrics interesting for the user, such as the speed of the racket just before impact with the ball in the case of tennis, the height of a jump in the case of riding or skiing, or the acceleration at the moment of impact in the case of boxing. Thus, at the output of the analyzer module, a flow of movements or sports gestures is available, each movement being identified, dated and characterized by metrics adapted to the sport practiced.

During a statistical analysis step E12, the server 5 performs statistical processing of all or part of the raw data contained in the database 7 to improve motion recognition algorithms. For this purpose, the raw data is processed by the server 5 according to statistical analysis methods, so as to provide one or more data sensors:

more appropriate algorithm configuration settings; or an improved version of the algorithm.

Depending on the particular gestures of an athlete for a given sport, the configuration of the motion recognition algorithm implemented by one or more sensors may be such that the classification reliability of the movements is not optimal.

In this case, the raw data provided by the sensor of this sportsman and recorded in the remote database during a training session are used by the data server 5 to adapt the classification algorithm of the classifier and consequently the classifier implemented by the server.

The adaptation of this classifier makes it possible to improve the recognition reliability of the algorithm according to the statistical analysis of the raw data of a particular sportsman or a group of athletes, the algorithm for recognizing a sensor being derived from the classifier implemented on the server.

Thus, the classification algorithm can be adapted or reconfigured according to specific data of the athlete. This adaptation is performed according to known statistical analysis methods for customizing the recognition algorithm, for example of the C4.5 type.

Statistical analysis methods advantageously allow the consideration of a very large number of individual data that could not be processed manually by a human being. This is particularly the case for improving the recognition reliability of a generic algorithm for a sport practiced by a large number of users.

As an illustrative example, the statistical analysis methods can exploit the Hamming distance criterion to reconfigure or adapt the classification algorithm of the classifier module implemented on the server. Thus, the motion recognition algorithm is reconfigured or adapted, so that the Hamming distance is as large as possible to limit the risk of confusing movements. The adaptation of the classification algorithm of the classifier module according to the Hamming distance by statistical analysis methods constitutes a means for determining configuration parameters of the algorithm making it possible, for example, to maximize the Hamming distance. for an optimal recognition of movements.

As an example, it is assumed that in the motion recognition algorithm initially implemented on the sensors, only two qualifiers (Qualifiers) are used and that the minimum Hamming distance observed for the entire population of athletes is equal to 0.25 (normalized Hamming distance). If a particular athlete (ie not included in the initial population) has a Hamming distance equal to 0, it is necessary to try to improve the conditions of recognition of the actions of this athlete.

As part of boxing, the inventors initially found that the Hamming distance was 0.25 considering the following two Qualifiers: "Variation of the acceleration along the local axis X" and "Variance of the rotational speed around the local axis Z ". Then by performing more complete tests with new types of athletes, this distance is increased to 0 indicating a confusion for the recognition of certain gestures, this being due to a more pronounced movement of the rotation around the local axis X.

In this case, the correction or adaptation of the algorithm consists in replacing the

Previous qualifiers by the "Variation of the acceleration in the terrestrial fixed plane X, Y" and the "Variance of the speed of rotation around the terrestrial fixed axis Z", in order to see increase the normalized Hamming distance for the all athletes (including the particular athlete) to a value equal to 0.3. This is typically the kind of situation that can occur for a new athlete with a very particular gesture.

All or part of the raw data stored in the remote database can be presented as a cloud of measurement points on a two-dimensional diagram, as shown in Figure 5a, where the x-axis represents the variance the mean linear acceleration along the local axis X (Vyx) and the ordinate axis represents the rms value of the rotation speed around the local Z axis (Vooz). According to this representation, a first cloud of points Gl is centered on a first coordinate reference point A (Vyx (A), Vcoz (A)) designating the center of gravity relative to a first gesture, while a second point cloud G2 is centered on a second coordinate reference point B ((Vyx (B), Vcoz (B)) designating the center of gravity relative to a second gesture.

The Hamming distance observed for a population of tested athletes is illustrated by the vector dl. In this representation, the standardized Hamming distance is used. To express the normalized distances, the values Va-y are divided along the ordinate axis by the value of the ordinate Vcoz (A) of the first reference point A and the values Vyx are divided along the abscissa axis by the value of the Vyx abscissa (B) of the second reference point B. The normalization facilitates the comparison of performances of different algorithms.

At the end of the E12 analysis of the raw data collected, the motion recognition algorithm used by the data sensor (s) that provided the raw data is adapted, so as to maximize the characterizing Hamming distance. the signature difference of each movement.

For example, we are interested in the Hamming distance between two types of detected movements characterized by two parameters, such as the average of the acceleration and the rms value of the rotation speed around an axis.

In this example, the statistical analysis method implemented by the server 7 comprises:

a partitioning step which consists in defining, on the representation of FIG. 5a, three disjoint zones SI, S2, S3 according to the example illustrated in FIG. 5b: a first zone SI corresponding to the first gesture, a second zone S2 corresponding to the second gesture, and a third intermediate zone S3 located between the first and second point clouds; - A test step which consists in checking whether each of the measuring points belongs to one of the three predefined zones.

In this example, partitioning is defined as follows:

the first zone SI is defined as the intersection of two straight lines x1, x2 parallel to the ordinate axis and two straight lines y3, y4 parallel to the abscissa axis; the second zone S2 is defined as the intersection of two straight lines x3, x4 parallel to the ordinate axis and two straight lines y1, y2 parallel to the abscissa axis;

the third zone S3 is defined by the area between a first line NI tangential to the first cloud of points G1 and passing through the reference point O formed by the intersection of the abscissa axis and the axis ordinates, and secondly a second straight line N2 tangential to the second cloud of points G2 passing through the landmark O, these two lines forming a minimum angle between them.

In the example described above, the configuration parameters of the motion recognition algorithm are as follows:

the variance of the acceleration in the terrestrial fixed plane X, Y (Vyx, _y ) and the variance of the speed of rotation around the Z axis (Vcoz) being selected as qualifiers (qualifying or qualifying values) adapted, the parameters x1, x2, y3, y4 defining the first area S1, the parameters x3, x4, y1, y2 defining the second area S2, and the two tangent lines N1, N2 defining the third area S3.

Depending on the positioning of the values measured by the accelerometer and the gyrometer on these areas, we can distinguish the movements between them. For the sake of simplicity, only two qualifiers (qualifier values) have been used in the example above. However, in practice, we can have five or six qualifiers to characterize each movement.

A user with movements whose recognition leads too often to near points (i.e. low Hamming distance, for example equal to 0) will be considered less reliable than a person producing movements with a high Hamming distance. A low Hamming distance is an indicator of reduced recognition reliability.

The statistical analysis of the collected raw data leads to increasing the Hamming distance, which has the effect of making the motion recognition algorithm performed by the motion recognition algorithm more reliable. Statistical analysis algorithms may be powered by more qualifier values that potentially increase the Hamming distance to further improve recognition reliability.

Based on the results of the statistical analysis performed on the collected raw data, it is determined whether improvements can be made to the motion recognition algorithm or the data sensor firmware including the recognition algorithm.

In the case of supervised learning, the adaptation of the classifier at the server level can be assisted by a human being who defines in advance the types of movement considered for the statistical analysis. If the movement has been previously classified by the user, the classifier module implemented at the server can operate autonomously and the firmware can be modified or generated automatically. Human intervention is planned to check the correct functioning and to authorize the official diffusion of the modified or generated firmware.

Subsequently, the application of the data sensor, this application containing one or more motion recognition algorithms, and other functions such as battery charge management, display management, will be designated by firmware. , Bluetooth communication.

It is assumed that the raw data come from the same data sensor used by the same sportsman for the motion recognition of a given sport. In this case, the statistical analysis step advantageously makes it possible to extract configuration parameters intended to configure the application executed on the data sensor according to the specificities of the sensor and / or the sportsman concerned.

Assuming that the raw data is derived from a plurality of data sensors used by different sportsmen for the motion recognition of a given sport, the statistical analysis step advantageously makes it possible to extract useful update parameters. to evolve the deployed application on all the sensors concerned.

During an updating step E ₁₄ , a new version of firmware is generated for one or more sensors requiring such an update, according to the results of the statistical analysis step E12, this generation being able to be implementation implemented by the server, without or with the intervention of a human being especially for control or verification purposes.

By way of illustrative example, the value of the Hamming distance resulting from the implementation of the classification algorithm of the classifier module by the server 5 constitutes a result making it possible to decide on a modification of the firmware. Thus, an update can be performed according to the Hamming distance calculated during the statistical analysis stage relating to all or part of the data stored in the database. In this sense, the Hamming distance is a parameter that potentially triggers a firmware update.

In case of modification, the new firmware is sent by the server 5 to the mobile phone 3 which transmits it to the (s) sensor (s) of data 1 concerned (s) by this update. The installation of the new firmware on the data sensor (s) can be controlled by the user using the application installed on his mobile phone.

The generation and the transmission of the new firmware by the server and the execution of its installation on the data sensor constitute the update step Ei4 of the application, this application being contained in said firmware.

Depending on the results of the statistical analysis, an update of the application executed on the data sensor is not essential to improve the recognition of movements.

It is then sufficient to reconfigure the application according to the configuration parameters obtained at the end of the statistical analysis step E ₁₂ . In this case, the configuration parameters are sent by the server 5 to the data sensor 1 via the mobile phone 3. Upon receipt of these parameters, the data sensor configures the motion recognition application according to the parameters received.

The obtaining and sending the configuration parameters and the application of these sensor on the application parameters constitute the configuring step E _15.

In a particular embodiment, the server 5 supplies to one or more data sensors a classification parameter of the movement, said parameter being determined according to analysis parameters, such as the Hamming distance.

In this case, the motion classification parameter constitutes a configuration parameter determined according to the results of the statistical analysis step E ₁₂ . Naturally, to meet specific needs, a person skilled in the field of the invention may apply modifications to the invention as described above without departing from the scope of the invention. For example, the motion recognition algorithm may be executed by the application installed on the mobile phone rather than by the application of the data sensor.

The present invention is not limited to the specific embodiments that have been described above, and the modifications that are within the scope of the present invention will be apparent to one skilled in the art.

Claims

A data processing method for improving the motion recognition implemented by an application executed on at least one data sensor (1;) within a data processing system, said method being characterized in that It includes the following steps by the data processing system:

collecting (E ₄ ) raw data by said at least one sensor (1);

storing (Eio) raw data in a remote database (7); - analyzing (E ₁₂ ) all or part of the raw data stored in the remote database (7); and

parameterizing (E ₁₄ ) and / or updating (E ₁₅ ) of said application as a function of the analysis results obtained at the end of the analysis step (E ₁₂ ); and that it further comprises:

a preliminary test step (E ₂ ), during which a criterion of reliability of motion recognition is evaluated when a movement is recognized by the application, so that the other steps of the method are initiated as a function of the value of the reliability criterion with respect to a predetermined threshold.

2. Method according to claim 1, characterized in that an identifier (I) qualifying the type of movement performed by a user of said at least one sensor is associated with said raw data (D) during the collection step (E ₄ ), the identifier being manually entered by the user or automatically according to a predefined sequence of movements to be carried out, the identifier (I) and the data (D) being transmitted to said database (7).

3. Method according to any one of claims 1 to 2, characterized in that said collected raw data (D) are sent by said at least one data sensor (1) to a mobile terminal (3) adapted to transmit said data (D) to the remote database (7).

4. Method according to any one of claims 1 to 3, characterized in that said raw data (D) comprise inertial data of acceleration and / or speed.

5. Method according to any one of claims 1 to 4, characterized in that it further comprises a packaging step (Es), during which the raw data collected (D) are packaged, so as to provide a calibrated data structure.

6. Method according to any one of claims 1 to 5, characterized in that the parameterizing step (E14) includes supplying said at least one data sensor (1;) with at least one configuration parameter determined in function of the results of said analysis (E12).

7. Method according to any one of claims 1 to 6, characterized in that the updating step (E ₁₄ ) includes supplying said at least one sensor (1;) with an application updated according to the results of said analysis (En).

A data processing method for improving the motion recognition implemented by an application running on at least one data sensor (1;) within a data processing system, said method comprising the following steps by said sensor:

collecting raw data (E ₄ ) by said at least one sensor (1; 1 ');

sending the collected raw data to a remote database (7); and receiving at least one configuration parameter and / or an update of said application, said at least one configuration parameter and / or said update being determined according to the results of an analysis (E12) of any or part of the raw data stored in the remote database (7); and that it further comprises:

a preliminary test step (E ₂ ), during which a criterion of reliability of motion recognition is evaluated when a movement is recognized by the application, so that the other steps of the method are initiated according to the value of the reliability criterion with respect to a predetermined threshold.

A data processing system for improving motion recognition implemented by an application associated with at least one data sensor (1;), said system comprising:

means for collecting raw data by said at least one sensor; storage means for storing raw data in a remote database (7);

means for analyzing all or part of the raw data stored in the database (7); and

means for parameterizing and / or updating said application as a function of the results obtained by said analysis means.

A data sensor adapted to enhance motion recognition and comprising motion recognition means, said sensor characterized by further comprising:

means for collecting raw data;

means for sending said data to a remote database (7);

means for receiving at least one determined configuration parameter and / or said adapted application according to the results of an analysis of all or parts of the raw data stored on the remote data base (7).

11. A data server for improving the motion recognition implemented by an application associated with at least one data sensor (1;), said server comprising:

means for storing said raw data in a database (7); means for analyzing all or part of the data contained in said database; and

12. Computer program comprising instructions adapted to implement the method according to any one of claims 1 to 10 when said program is run on a computer.

13. An information storage medium, removable or not, partially or completely readable by a computer or a microprocessor comprising code instructions of a computer program for the execution of each of the steps of the method according to any one Claims 1 to 10.