Data processing method, method for determining a critical exertion intensity for an animal, and associated system

A data processing method analyzes acceleration data to determine a critical effort intensity for animals, addressing the limitations of existing methods by providing a reliable intensity threshold without laboratory tests, suitable for personalized activity classification and natural settings.

WO2026234059A1PCT designated stage Publication Date: 2026-11-12UNIV SAVOIE MONT BLANC +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/EP2026/063036
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-09
Filing Date
2026-05-06
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Existing methods for determining individual exercise intensity thresholds, such as metabolic equivalent (MET) classification and laboratory-based incremental tests, are not suitable for personalized activity classification due to their reliance on physiological data and laboratory settings, making them impractical for widespread use, especially in the study of non-human animals.

Method used

A data processing method using a server device to analyze acceleration data from an animal, calculating moving averages and fitting a decreasing and convergent model to determine a critical effort intensity without requiring blood tests or laboratory settings, allowing for the determination of an intensity threshold that an individual or animal can sustain without triggering significant fatigue.

Benefits of technology

Enables the determination of a critical effort intensity suitable for individual activity classification, applicable in natural settings, without the need for laboratory tests, facilitating the study of spontaneous animal behavior and providing a reliable intensity threshold for moderate and vigorous activities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure EP2026063036_12112026_PF_FP_ABST
    Figure EP2026063036_12112026_PF_FP_ABST
Patent Text Reader

Abstract

The invention relates to a data processing method, implemented by a server device, for obtaining a critical exertion intensity (DBAc) for an animal, the processing method comprising, on the basis of a time series of data representative of at least one acceleration generated by the animal, and on the basis of a plurality of time windows, having durations that are distinct from one another: - calculating sliding averages of the time series by virtue of the plurality of time windows; - determining at least one maximum (Smax(Δt1), Smax(Δt2)), referred to as "record", of each of the calculated sliding averages; and - adjusting a decreasing and convergent model (Smod(Δt)) according to the determined records, the model being a function of the distinct durations (Δt1, Δt2) of the time windows used to calculate the sliding averages; - determining an asymptote (DBAc) of the adjusted model (Smod(Δt)), the asymptote (DBAc) being the critical exertion intensity of the animal.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] "DATA PROCESSING METHOD, METHOD FOR DETERMINING CRITICAL EFFORT INTENSITY FOR AN ANIMAL AND ASSOCIATED SYSTEM"

[0002] TECHNICAL FIELD

[0003] The technical scope of this application relates to the determination of a critical effort intensity for an animal, whether human or non-human, and more particularly in a way that does not require dedicated laboratory facilities.

[0004] STATE OF THE ART

[0005] In 2020, the World Health Organization (WHO) recommended at least 150 to 300 minutes of moderate-intensity aerobic physical activity, or at least 75 to 150 minutes of vigorous-intensity aerobic physical activity, or an equivalent combination of moderate- and vigorous-intensity activity throughout the week, to achieve substantial health benefits. One of the main challenges lies in characterizing the intensity of activities. The WHO recommendations include non-exhaustive lists of activities classified as moderate or vigorous. Moderate activities include, for example, brisk walking, household chores such as cleaning, or low-intensity cycling. Vigorous activities include, for example, hiking, jogging, or intense cycling. These activities are provided as guidelines and do not necessarily correspond to physiological reality.

[0006] Moderate and vigorous activities can also be classified according to a metabolic equivalent of the activity, the unit of which is "MET," which corresponds to oxygen consumption at rest (this basal metabolic rate is considered to be around 3.5 ml / min / kg). Sedentary activity has a metabolic equivalent between 1.0 MET and 1.5 MET. Light-intensity activity has a metabolic equivalent between 1.6 MET and 2.9 MET. Moderate-intensity activity has a metabolic equivalent between 3 MET and 5.9 MET. Finally, vigorous-intensity activity has a metabolic equivalent of 6 MET or more. However, oxygen consumption varies according to age, sex, and fitness level.Therefore, the absolute thresholds of 3 MET and 6 MET for classifying moderate and vigorous activities are not appropriate for properly characterizing individual relative intensity.

[0007] Therefore, there is a need to reliably determine an intensity threshold for an individual in order to create a personalized activity classification for that person. This need can be applied more generally to the animal world in the broadest sense, encompassing all non-human individuals.

[0008] Laboratory methods exist to determine individual intensity thresholds that separate exercise ranges (light, moderate, vigorous). One example is the so-called "incremental" method, which involves progressively increasing an individual's walking speed until a sharp change in a physiological variable is observed, indicating a shift in exercise range. Examples of such physiological variables include blood lactate levels (referred to as "MLSS," for "Maximum Lactate Steady State"), respiratory rate, or cardiac drift. However, this type of measurement requires physiological data, such as one or more blood samples taken during incremental exercise (where the intensity of the effort increases gradually at regular intervals, for example, every few minutes), as well as laboratory analysis of the blood. Therefore, this method is difficult to implement outside of a laboratory setting and on a large scale.Moreover, this method, when applied to the animal world, is even more difficult to implement for wild animals.

[0009] SUMMARY

[0010] A first aspect of the invention relates to a data processing method, implemented by a server device, to obtain a critical effort intensity for an animal, the processing method comprising the steps of:

[0011] receive, from a client device, at least one initial time series of data representative of at least one acceleration generated by the animal for several instants over time;

[0012] from a plurality of time windows, presenting distinct durations from each other, calculate, for each of the time windows of the plurality of time windows, a convolution product of the first time series of data with said time window, called "moving average";

[0013] determine at least one maximum, called a "record", of each of the calculated moving averages; and fit a decreasing and convergent model on the determined records, said model being a function of the distinct durations of the time windows used to calculate the moving averages;

[0014] determine an asymptote of the fitted model, said asymptote being the critical effort intensity of the animal.

[0015] The invention offers a solution, at least partial, to the aforementioned problems. In particular, it allows for the determination of a critical effort intensity that corresponds to an intensity threshold. This is an effort intensity that an individual or animal is capable of sustaining without triggering significant acute fatigue. In other words, it corresponds to the most intense moderate activity that the individual or animal is capable of sustaining.

[0016] The invention has the major advantage of not requiring blood tests or other physiological measurements of activities in a constrained setting, such as a laboratory, to obtain a critical effort intensity. The invention can, for example, be implemented partially or even entirely in nature. Therefore, the invention is particularly well-suited to the study of spontaneous animal behavior (including human and non-human).

[0017] A second aspect of the invention relates to a method for determining a critical intensity for an animal, implemented by a customer device, the determination method comprising the steps of:

[0018] To acquire, using an accelerometer, data representative of an initial acceleration generated by the animal over several instants in time, in order to form a first time series of data; and

[0019] Send the first time series of data to a server device configured to implement the processing method according to the invention, so as to obtain the critical effort intensity of the animal.

[0020] The third, fourth and fifth aspects of the invention relate respectively to: a server device comprising means for receiving at least a first time series of data from a client device, and a computer configured to implement the steps of the processing method according to the invention;

[0021] a client device comprising means for sending a first time series of data to a server device, as well as an accelerometer and a computer configured to implement the steps of the determination process according to the invention; and

[0022] a client-server system comprising a server device according to the invention and a client device according to the invention.

[0023] The sixth and seventh aspects of the invention relate respectively to:

[0024] a computer program comprising instructions which, when the program is executed by the server device according to the invention, cause the server device to receive at least a first time series of data from a client device and to implement the steps of the processing method according to the invention; and

[0025] a computer program comprising instructions which, when the program is executed by a client device, cause the client device to send at least a first time series of data to a server device and to implement the steps of the determination process according to the invention.

[0026] The eighth and ninth aspects of the invention relate to non-volatile and computer-readable storage media on which the aforementioned computer programs are respectively recorded.

[0027] BRIEF DESCRIPTION OF THE FIGURES

[0028] The aims, objects, features and advantages of the invention will become clearer from the detailed description of an embodiment thereof, which is illustrated by the following accompanying drawings in which:

[0029] Figures 1 and 2 schematically represent two embodiments of a treatment process according to the invention, as well as one embodiment of a determination process according to the invention. These figures also schematically represent a system for implementing the aforementioned processes.

[0030] Figure 3 schematically represents two examples of data processing such as can be implemented by the processing method according to the invention.

[0031] Figure 4 schematically represents another example of data processing such as can be implemented by the processing method according to the invention.

[0032] Figure 5 schematically illustrates an example of determining a maximum force intensity that can be implemented using the determination method according to the invention. The drawings are provided by way of example and are not intended to limit the scope of the invention. They are schematic representations of principle intended to facilitate understanding of the invention and are not necessarily to scale with practical applications. Furthermore, they are not necessarily representative of reality.

[0033] DETAILED DESCRIPTION

[0034] Before beginning a detailed review of embodiments of the invention, optional features that may be used in combination or alternatively are stated below.

[0035] In one embodiment of the treatment process, the fitting of the asymptotic model is carried out by quantile regression with a quantile greater than or equal to 75%.

[0036] In one embodiment, the treatment process further includes the steps of:

[0037] receive a second time series representing at least one altitude of the animal over time, the second time series being acquired at least partly simultaneously with the first time series;

[0038] Prior to calculating moving averages, correct the first time series using the second time series, each instant of the corrected first time series being a function of the time derivative of the second time series at the corresponding instant.

[0039] In one embodiment of the processing method, the second time series is representative of at least one barometric data point or geolocation data point of the animal. In one embodiment, the processing method further includes the steps of: receiving a third time series representative of at least one geographical data point of the animal over time, the third time series being acquired at least partly simultaneously with the first time series;

[0040] the correction of the first time series being further a function of the third time series, each instant of the first corrected time series also being a function of the time derivative of the third time series at the corresponding instant.

[0041] In one embodiment of the treatment process, the first time series represents a vector sum of the acceleration generated by the animal as a function of time.

[0042] In one embodiment of the treatment process, the acceleration generated by the animal as a function of time is sampled at a sampling frequency greater than or equal to 1 Hz.

[0043] In one embodiment, the processing method includes a step, carried out prior to the calculation of moving averages, of correcting the first time series to remove the effect of gravity experienced by the individual or animal over time.

[0044] In one embodiment of the processing method, the duration of each time window is between five seconds and two hours, and preferably between ten seconds and thirty minutes.

[0045] In one embodiment, the determination process includes the additional steps of:

[0046] to acquire, using the accelerometer, data relating to a second acceleration generated by the animal during an activity, for several instants over time; to compare, for at least one instant, the data relating to the second acceleration acquired with the critical effort intensity of the animal in order to obtain a category of the activity; and

[0047] indicate the result of the comparison.

[0048] In one embodiment of the determination method, when the data relating to the second acceleration is less than the critical effort intensity, indicate information that the activity is in a moderate regime; and when the second acceleration is greater than the critical effort intensity, indicate information that the activity is in a vigorous regime.

[0049] By "gravity," we mean a gravitational acceleration, measured at the Earth's surface, which is approximately 9.81 m / s². 2 .

[0050] By "animal" we mean an individual from the animal kingdom, including both human and non-human individuals.

[0051] By "interaction between two processes," we mean that there is a transmission of data between the two processes, either unidirectionally, from one process to the other, or bidirectionally. By "data or a series of data representing an analog quantity" or "data or a series of data relating to an analog quantity," we mean a result of the conversion or transformation of the analog quantity into digital data. The conversion or transformation of the analog quantity may include a step of formatting the analog quantity (such as scaling) or a step of formatting the digital result (such as scaling and / or vector calculation).

[0052] The various aspects of the invention will now be described with reference to the figures. Figures 1 and 2 schematically represent two embodiments of a processing method 1. This processing method 1 makes it possible to obtain a critical effort intensity DBAc (for critical Dynamic Body Acceleration) for an animal from data relating to an acceleration generated by the animal.

[0053] Processing method 1 can interact with a method for determining the critical intensity DBAc of the animal 2. Figure 1 schematically represents one embodiment of the determination method 2 as well as the interactions between the two methods 1 and 2 (i.e., the exchange of data between the two methods).

[0054] These two figures illustrate two embodiments of a system for implementing the two aforementioned processes. This system operates on a client-server principle. Specifically, it comprises a server device 3 and a client device 4. A "server" or "server device" is defined as a device configured to provide a service in response to a request, which may correspond to the reception of data. A "client" or "client device" is defined as a device configured to communicate with the server.

[0055] In the context of the invention, client 4 is intended to acquire data and send it, either raw or after preliminary processing, to server 3 for further processing. In Figure 1, client 4 is also configured to interact with a user, for example, to present a result to a user. In Figure 2, the system also includes an additional client 5 for, for example, informing a user or a third party.

[0056] Server 3 is intended to process the data sent by client 4. It can also be intended to send a processing result to client 4 or to additional client 5 to display this result to the user or third party.

[0057] To communicate with client 4, server 3 includes, for example, a network interface, allowing it to receive and / or send data, respectively, to or from a computer network. The computer network can be a private or public network. When accessible via a network such as the internet, server 3 can constitute a cloud computing service.

[0058] Alternatively or in combination, the communication means of server 3 may include a serial communication channel. These means include, for example, an electronic connector to accommodate such communication. This is, for example, a "USB" (for "Universal Serial Bus") interface, "l 2Server 3 also includes computing resources to process the data it may receive. These resources can be generic computing resources such as a calculator, a programmable logic circuit, a computer, or a supercomputer.

[0059] Client 4 is configured to interact with server 3. It also includes means 42 enabling it to communicate with server 3. These means 42 include, for example, a network interface allowing it to send and / or receive data to and from a computer network. In this way, client 4 can send or receive data to and from server 3. The communication means 42 may also include a serial communication channel implementing, for example, USB, I2C, RS-232, RS-422, RS-485, or equivalent communication.

[0060] Client 4 may also include computing resources enabling it to perform preliminary processing on the data before sending it to server 3. These computing resources can be generic computing resources such as a calculator, a programmable logic circuit, a computer, or even a supercomputer. However, they are advantageously compact so that the animal being studied can carry Client 4 with it. Client 4 is, for example, a so-called "connected" object, that is, one capable of connecting to a network. Examples of Client 4 include a smartphone, a smartwatch, or an actigraph. Client 4 can also be a tracking tag or a geolocation beacon that can be attached to the animal.

[0061] Client 4 also includes an accelerometer 41. This accelerometer measures the acceleration generated by at least one part of the animal. This part can be an upper or lower limb, such as an arm, leg, or paw. Client 4 therefore allows for the measurement of the acceleration of this part during various activities performed by the animal. It can be attached to the part using a removable attachment such as a bracelet, collar, or adhesive strip. For example, when the accelerometer is attached to the animal's wrist, it measures the swinging motion of the arm while walking. It also measures more intense accelerations during activities that heavily involve the arms and hands, such as housework or sports. The accelerometer 41 can also be attached to the animal's pelvis or neck.Thus, the accelerometer makes it possible to measure the acceleration resulting from the movement of the animal, regardless of the type of activity it is performing.

[0062] The steps of processing method 1 and determination method 2 can be carried out using different infrastructures, for example, one infrastructure for server 3 and another for client 4. Indeed, for certain steps, it is not necessary to acquire data or be near the animal. In other words, these steps can be performed remotely to take advantage of greater computing power than that offered by client 4's infrastructure. However, it is also possible for client 4 and server 3 to be part of the same infrastructure. This infrastructure could be, for example, a smartphone, a smartwatch, or an actigraph worn by the animal. In this case, client 4 and server 3 would be, for example, computer modules running on this single infrastructure.This method of implementation has the advantage of being able to carry out all the processes described below at the location where the data acquisitions are carried out, for example in natura.

[0063] An embodiment in which the client 4 and the server 3 are run by separate infrastructures (for example, as illustrated in Figures 1 and 2) reduces the energy requirements of the infrastructure corresponding to client 4. This infrastructure can then benefit from extended battery life and / or reduced weight and / or improved mechanical robustness. These advantages can be further enhanced when the system includes an additional client device 5, as illustrated in Figure 2. The processing or display functions can thus be entrusted to another terminal. In the embodiment of Figure 2, only the most essential functions of acquiring and sending data related to acceleration are performed. Thus, energy consumption and size can be minimized.When client 4 and server 3 are part of different infrastructures, client 4 then advantageously includes a means of data storage such as computer memory, allowing it to record data acquired during different efforts or activities carried out by the individual and to be able to transmit them subsequently to server 3.

[0064] The steps of the determination process 2 will now be described in detail. These steps are preferably implemented by customer 4. However, some of the steps of the determination process 2 can be implemented by the additional customer device 5 as illustrated in Figure 2.

[0065] In the implementations shown in Figures 1 and 2, the determination method 2 comprises an acquisition 21, using the accelerometer 41 of the client 4, of a data point A relating to an acceleration generated by the animal, at several instants during time t. The data point A corresponds, for example, to a measurement of a signal from the accelerometer 41 when the latter is worn by the animal. By "data point" relating to an acceleration, we mean, for example, a digital data point that corresponds to the analog-to-digital conversion of a signal from an accelerometer.

[0066] In its simplest implementation, as illustrated in Figure 2, the acceleration data A is a vector and corresponds to the Ax, Ay, and Az components of the acceleration generated by the animal. This complex data A can then be transmitted. It has the advantage of not requiring costly formatting calculations.

[0067] In a development, for example illustrated in Figure 1, the data A relating to acceleration corresponds to a vector sum of the components Ax, Ay, Az of the acceleration generated by the animal. Note that a vector magnitude is a vector sum. This implementation method requires calculating the vector sum. However, it allows for the transmission of a smaller data file to server 3.

[0068] Before its analog-to-digital conversion, the acceleration may have undergone preliminary processing, such as amplification and / or filtering. In Figure 1, the data point A relating to the acceleration generated by the animal corresponds to a vector sum of the Ax, Ay, and Az components of the acceleration. The measured data is then processed, preferably by client 4, to form a first time series S of data. By "time series of data," we mean a set of digital data, each digital data point corresponding to a time (for example, an acquisition time), with the timestamps arranged in ascending order. In other words, the first time series S corresponds to a sampling of the signal from the accelerometer.

[0069] The acceleration acquisition sampling can be regular or irregular. The sampling frequency can be greater than 1 Hz, for example, between 1 Hz and 100 Hz.

[0070] In a second step, the determination process 2 includes sending the first time series S to the server 3 so that it can calculate a critical effort intensity DBAc associated with the animal.

[0071] Thus, by simply measuring the acceleration of a part of an individual, it is possible to determine a critical effort intensity for that particular animal. From this, it becomes possible to classify the activities that this individual is likely to perform as moderate or vigorous. Therefore, there is no need for laboratory tests, blood samples, or any other biological sampling.

[0072] For robust determination, it is preferable that acceleration be measured at a body part that is stressed during both moderate and vigorous activities. This part could be, for example, an arm, wrist, or hand. It could also be a leg or ankle. Other possibilities include the pelvis, trunk, neck, or head. In one implementation, the determination method 2 may include preprocessing 22 of the acceleration data A before sending it to the server 3. For example, to limit the amount of information sent to the server, the determination method may involve calculating an algebraic sum or a vector sum, such as a vector norm, of the acceleration data A as a function of time. Indeed, acceleration is a vector variable. The vector sum (or vector norm) can be called "VeDBA".The algebraic sum of data A can be called "ODBA". Calculating a sum, for example a vector sum, thus allows the volume of data to be sent to be halved or even reduced by a factor of three. Furthermore, a scalar value, such as the vector sum of acceleration, is generally sufficient to adequately determine an individual's critical effort intensity. Therefore, client 4 can be configured to calculate this vector sum from the measured data. The first time series S thus corresponds to the vector sum of data A relating to the acceleration generated by the individual. Alternatively, the calculation of the vector sum, such as the magnitude, can be performed using a subset of the acceleration components. Considering the acceleration projected onto a terrestrial geodetic system (such as the "WGS 84" system), the components considered for calculating the vector sum can be the components tangent to this system.Advantageously, the determination method 2 can, before sending the time series S to the server 3, correct the data used to form the time series S in order to remove the effect of gravity on the time series S. This correction is achieved, for example, by filtering each component Ax, Ay, Az of the data A relating to an acceleration to remove a DC component. This filtering is, for example, a high-pass filter with a low cutoff frequency greater than or equal to 0.01 Hz. Alternatively, this filtering can be performed by calculating a moving average of each component Ax, Ay, Az of the measured acceleration and subtracting each moving average of the associated Ax, Ay, Az component. The moving average can be calculated from a time window with a duration between 1 s and 10 s, for example, 2 s.The determination process 2 may include additional steps, one of the objectives of which is to inform a user. The user may be the individual for whom the critical intensity is being measured. Alternatively, it may be a third party, for example, a doctor or an ecologist. These additional steps may be implemented by the client 4 itself, as illustrated in Figure 1, or at least partially by the client 4 and an additional client device 5. This implementation mode corresponds, for example, to Figure 2. In one implementation mode, the determination process 2 may include a receiving step 25 of a critical effort intensity DBAc, for example, sent by the server 3. This reception 25 may be carried out by the client 4 (as in Figure 1). This is particularly relevant when the critical intensity result is communicated directly to the individual.Alternatively, this reception 25 can be carried out by the additional client 5 (case in Figure 2). This is particularly relevant when the critical intensity result needs to be consulted by a third party.

[0073] To display information, client 4 or additional client 5 may include a display means, such as a screen. This screen may be used to display the critical effort intensity DBAc.

[0074] The critical effort intensity DBAc can also be used to classify activities performed by the animal. For example, the determination process 2 may include an additional step consisting of acquiring 24 an additional time series of data S2 relating to the acceleration generated by the individual during an activity. This acquisition 24 may be performed before or after the determination of the critical effort intensity DBAc. It is preferably performed by client 4 in order to utilize the accelerometer. Alternatively, the additional client 5 may also include an accelerometer to perform the acquisition of this additional time series S2.

[0075] Method 2 includes a step 26 comparing this additional time series S2 with the resulting critical effort intensity DBAc. This allows the activity for which the additional time series was acquired to be classified into an activity category, such as "moderate activity" or "vigorous activity." For example, if the acquired additional time series S2 is less than the critical effort intensity DBAc, the activity can be categorized as "moderate." However, the range of moderate activity can be restricted to activities for which the acceleration (and therefore the resulting time series S) is greater than 0.26g. In this way, rest periods are not classified as moderate activities. If the additional time series S2 is greater than the critical effort intensity DBAc, then the activity can be categorized as "vigorous."The result of the comparison can be indicated to the user, for example by means of the display of the client device or by means of another display.

[0076] It is also possible to determine a data relating to an individual's state of fatigue, corresponding for example to the removal of the critical effort intensity value DBAc from the additional time series S2. The integral of this new signal then indicates the individual's state of fatigue as a function of time.

[0077] In addition, comparison step 25 can also calculate one or more durations, for example daily durations, corresponding to one or more sums of durations of activities categorized as moderate and / or vigorous (e.g., DM is the sum of moderate activity durations and DV is the sum of vigorous activity durations). Comparison step 26 can also include a comparison of this or these durations with a threshold duration, such as a recommended duration. For example, the WHO recommends that a physical activity duration, calculated as:

[0078] APS = (DM + 2*DV) / 300

[0079] is greater than or equal to a threshold duration of 150 minutes per day for an individual between 18 and 64 years old.

[0080] The DM, DV and / or APS durations can be indicated to the user, for example by means of the client display 4.

[0081] The steps of processing method 1, which can be performed at least partially by server 3, will now be described in detail. Processing method 1 aims to process the first time series S of data (also simply called the "time series") that can be sent by client 4. The first time series S can also be sent by another device. The first time series S represents an acceleration generated by the animal whose critical effort intensity is to be determined.

[0082] Figures 3 and 4 schematically represent the processing steps applied to the first time series S. Initially, and as illustrated in Figure 3, a plurality of moving averages <s>(At1), <s>(At2) is calculated from the received time series S. Each moving average <s>(At1), <s>(At2) is calculated using a time window. A "time window" is defined as a part-continuous function that is non-zero at a central point and zero everywhere else. The central point has a duration, denoted At. At the central point, the continuous function can be constant. It can also have a different form, for example, symmetric, asymmetric, or periodic.

[0083] The plurality of moving averages <s>(At1), <s>(At2) is obtained for time windows of different durations At1, At2. A moving average is obtained by convolving the time series S with a time window. This operation smooths the data, to a greater or lesser extent depending on whether the duration of the time window is long or, conversely, short.

[0084] Figure 3 shows an example of a raw time series S, as a function of time t, represented by a solid line. This raw time series S exhibits four phases 1, 2, 3, and 4, corresponding, for example, to different activities, some rather moderate (low S values, in arbitrary units) and some rather sustained (high S values). This example also shows variation due to measurement noise and activity variation within each phase. The time series S was measured over a period of approximately sixty minutes. The two dashed curves <s>(At1), <s>(At2) correspond to moving averages calculated from time windows of different durations. The first curve <s>(At1) is calculated from a short time window At1, in this example, approximately nine minutes. The second curve <s>(At2) corresponds to a larger time window At2, in this example, of approximately nineteen minutes. The second curve <s>(At2) corresponds to a greater smoothing of the time series S.

[0085] The treatment process 1 further includes a step 15 for determining at least one maximum Smax(At1), Smax(At2), referred to as the "record", for each moving average <s>(At1), <s>(At2). Only a maximum value of the moving average is considered, and not the instant corresponding to that maximum value.

[0086] Figure 3 shows the recorded records Smax(At1) and Smax(At2) for the two moving averages. <s>(At1), <s>(At2). Figure 3 also shows a record SmaxO for the raw S time series.

[0087] Figure 4 illustrates an example where a plurality of records Smax(A1), Smax(A2) are reported as a function of the durations At1, At2 of the windows used to calculate the moving averages <s>(At1), <s>(At2). Each point in figure 4 therefore corresponds to a record Smax of a corresponding moving average.

[0088] In one embodiment, process 1 includes determining 15 several records Smax(At1), Smax(At2) for each moving average <s>(At1), <s>(At2). This improves the accuracy of determining the critical stress intensity DBAc. To achieve this, each moving average <s>(At1) is divided into several consecutive but non-overlapping segments. Each record Smax(At1) is determined for each segment of the moving average.

[0089] Each segment of the moving average can have a duration greater than the duration of the window At1 used to calculate the moving average. For example, in the example in Figure 3, where the durations At1 and At2 of the time windows are between 9 and 19 minutes, the segments can have a duration strictly greater than At2, for example, greater than or equal to 30 minutes. In practice, the time series S can have a total duration exceeding several days. The segments can then have a duration greater than 1 hour, for example, 4 hours, half a day, or even a full day.

[0090] Alternatively, the number of segments can be specified for each moving average. For example, it could be between 2 and 30, or 10. The duration of the segments is then adjusted so that they are consecutive and do not overlap between adjacent segments.

[0091] Processing method 2 includes a complementary step of fitting a decreasing and converging Smod(At) model to the Smax records as a function of the durations At of the time windows. This type of Smod model may have an asymptote that allows the critical effort intensity DBAc to be determined. The Smod model can be, for example, an inverse function, an exponential function, or a power function.

[0092] In some cases, this type of model can be called a "critical intensity model" because it allows the behavior of a system under stress to be described as a function of duration or number of repetitions. One of the remarkable aspects of the invention is applying this type of model to data relating to acceleration while considering the durations At of the time windows as significant parameters.

[0093] Indeed, using time windows of different durations, At1 and At2, to calculate the moving average allows us to consider different timeframes for the same data. Therefore, from a single data record, it is possible to isolate activities corresponding to different time scales.

[0094] Vigorous activity is detected for the moving averages. <s>(At1) calculated from short time windows, At1, because they are generally correlated with brief, high acceleration peaks. Moderate activity is detected using moving averages. <s>(At2) calculated from time windows of much longer At2 durations because they generally correspond to lower but continuous acceleration levels over a long period.

[0095] A model Smod(At) varying according to the duration At of the time window can be defined as follows:

[0096] Smod(At) = k / At + DBAc

[0097] where k is a constant and DBAc an asymptote of the model. As a reminder, the critical effort intensity corresponds to the level of intensity for an effort that the user is likely to sustain, without stopping, for a long duration. In other words, it corresponds to a long time window At, and by extension, tends towards infinity. We can therefore consider that the asymptote of the decreasing convergent model corresponds to the critical effort intensity DBAc. Thus, after fitting the aforementioned Smod(At) model to the Smax(At) records, we obtain the asymptote for At → ∞. The fitting method is of particular importance for obtaining an asymptote value. Indeed, there are several ways to fit a model to data. A frequently used method, for example, is least squares regression. This type of regression assigns an equal weight to all data. The effect is then to reduce the importance of data with a high value.The estimated asymptote therefore has a lower value. Figure 4 shows an example of a least-squares fitted Smod' model and the resulting DBAc' asymptote. The critical stress intensity is thus underestimated.

[0098] While this may have little effect on data with low variance, it can be a drawback when the data variance is high. When evaluating a critical effort intensity DBAc, it is best not to disregard data with high values, especially when the data variance is high.

[0099] To overcome these drawbacks, the fitting method used to adjust the Smod model to the Smax records advantageously implements a quantile regression with a quantile greater than or equal to 75%, preferably greater than or equal to 90%, and even more preferably greater than or equal to 99%, or even greater than 99.5%. In this way, the high records are more representative in the fitting of the critical Smod model, and the estimation of the critical intensity DBAc is therefore improved.

[0100] Processing method 1 may include an additional step of sending the determined critical intensity DBAc to a device such as customer 4 or additional customer 5.

[0101] Figure 2 shows that the time series S can be preprocessed, for example by server 3, before the calculation of moving averages.

[0102] In a first example, processing method 1 may include receiving a second time series Z representing, for example, the animal's altitude over time t. This second series Z is called "altitude." The altitude is acquired, for example, by client 4. This data may be barometric or derived from a geolocation and navigation satellite system (GNSS) data. For this purpose, client 4 is preferably equipped with a means of measuring atmospheric pressure and / or GNSS altitude. If it is geolocation data, it is preferably projected onto a model of the Earth's surface. The altitude is acquired at least partially simultaneously with the acquisition of the first time series S. However, the altitude may be acquired with a sampling frequency different from the sampling frequency used to acquire the individual's acceleration.The altitude Z, for example, is acquired with a sampling frequency between 0.01 Hz and 10 Hz.

[0103] Processing method 1 includes, prior to calculation 14, the moving averages <s>(At1), <s>(At2), a correction step 11 of the time series S using altitude Z, preferably in such a way as to take into account a slope i on which the individual is moving at the time the acceleration is measured. It is thus possible to determine if the animal is performing vigorous activity (for example, climbing stairs, a hill, or a mountain) even if the acceleration alone would suggest that the activity is rather moderate. The slope (i, as a percentage) is calculated at each instant of the time series S as:

[0104] AZ

[0105]

[0106] a ■ S(t~) ■ At

[0107] where a can be considered, by default, to be equal to 2.5. Thus, the corrected time series is, at each instant, equal to:

[0108] S CO rr(.t)=S(.t) ' P

[0109] where p is a correction factor which can be, for example, equal to:

[0110]

[0111] where a and b are coefficients that can be respectively equal to a = 41.6 and b = 14.9.

[0112] The coefficient 'a' is a slope relating an individual's acceleration to their average speed. While this coefficient is unique to each individual, it is generally equal to 2.5 ± 0.5. However, it is possible to determine 'a' experimentally to improve the determination of critical effort intensity. In this case, 'a' can be determined by calculating a slope between the individual's acceleration and their average speed. <vxy>In other words, it can be obtained by:

[0113] a = <ds dvxy>

[0114] In one embodiment, the processing method 1 may include receiving a third time series XY representing, for example, a geographic data point of the animal over time t. This third XY series is called "position". The position is, for example, acquired by the client 4. This is, for example, GNSS data, which can be acquired at least partially simultaneously with the altitude Z. The GNSS data advantageously takes the form of an ellipsoid of the Earth, considering a geodetic system such as "WGS84". The client 4 is preferably equipped with a means for measuring GNSS data. The XY position is acquired at least partially simultaneously with the acquisition of the animal's acceleration. The position can be acquired with a sampling frequency different from the sampling frequency used to acquire the time series S.The XY position is acquired, for example, with a sampling frequency between 0.01 Hz and 1 Hz.

[0115] The derivative of the position XY as a function of time thus allows us to calculate the velocity vxy of the animal as a function of time and simultaneously with the first time series S. In this way, we can determine the coefficient a according to the aforementioned relationship.

[0116] In the case where the position is acquired by client 4, it is possible to indicate to the user, for example by means of the display, the location in space and / or time of vigorous or moderate efforts as well as their temporality.

[0117] The maximum exertion capacity (DBAmax) of an individual or animal reflects their state of fatigue. It is a scalar quantity that depends on time. The maximum exertion capacity (DBAmax) can be described as "mobilizable" or even "theoretical," as it represents the limit that can be mobilized by the individual.

[0118] The critical effort intensity DBAc represents a threshold: beyond which (in the range we previously described as "vigorous"), the individual becomes fatigued; and below which (in the range we described as "moderate") the individual recovers. The critical effort intensity DBAc therefore constitutes a fatigue threshold.

[0119] When an individual is fully rested, for example at t = 0, they can sustain an effort intensity for a very short duration (corresponding, for example, to At → 0) that is equal to an absolute record intensity. This is, for example, SmaxO in Figure 3. In other words, when the subject is rested, DBAmax(t = 0) is, for example, equal to max(Smax(At → 0)), where the max(...) operator is the maximum operator. We denote by "DBAi" the initial value of the maximum effort capacity when the individual is fully rested. Thus, when the individual is fully rested, DBAmax is, at best, equal to DBAi. There are several methods for determining DBAi, which we will detail later.

[0120] As an individual becomes fatigued, that is, as they perform vigorous activities (where the instantaneous effort, proportional to the acceleration A, is strictly greater than DBAc), their maximum exertion capacity DBAmax© decreases. It can decrease to a minimum value equal to the critical intensity DBAc. Therefore, the maximum exertion capacity DBAmax lies between DBAc and DBAl. We can define a normalized instantaneous quantity q(t), called the "fatigue index" or simply "fatigue," as

[0121] DBAmax(t)-DBAc

[0122] 11 f DBAI-DBAc

[0123] The fatigue index q(t) is between 0 and 1. It is:

[0124] • equal to 1 when the individual is completely rested, that is to say when DBAmax = DBAi;

[0125] And

[0126] • equal to 0 when the individual is completely fatigued, i.e. when DBAmax = DBAc. In one embodiment, illustrated by Figure 5, the determination method 2 may also include calculating the maximum effort capacity DBAmax of the individual or animal and, preferably, presenting (e.g. by display) this maximum effort capacity to a user (e.g. the individual).

[0127] To achieve this, determination method 2 involves calculating the maximum effort capacity DBAmax for each time step, based on the time series of supplementary data S2 relating to the acceleration generated by the individual during an activity. For each time step t, it is first determined whether the activity performed is vigorous or moderate. To do this, each supplementary data point S2(t) at time step t is compared to DBAc. When S2(t) > DBAc, the activity is vigorous; otherwise, the activity is considered moderate.

[0128] When the activity is vigorous, the maximum effort intensity DBAmax(t) at time step t is calculated from:

[0129] 1 DBAmax(t)=DBAmax(tl)-- (S2(t)-DBAc)

[0130]

[0131] T

[0132] where DBAmax(t-1) is the maximum effort capacity at the previous time step t-1, and T is a fatigability constant. Thus, as the activity becomes more vigorous, i.e., S2(t) > DBAc, then DBAmax(t) decreases, showing that the individual becomes fatigued and less and less capable of performing vigorous activity.

[0133] When the activity is moderate, the maximum effort capacity DBAmax© at time step t is calculated from:

[0134] „ ,, „ (DBAi-DBAmax(j>Ÿj\ , „ , DBAmax(t)=DBAmax(tl)+RI - gg^- gg^ - ) (DBAc-S2(t))

[0135]

[0136] where DBAi is the initial maximum effort capacity, and R is a recovery constant. Thus, as the activity is moderate, i.e., S2(t) < DBAc, then DBAmax© increases, showing that the individual rests and becomes capable of performing more vigorous activity again.

[0137] DBAmax© remains, of course, between DBAc and DBAl. The fatigue index q(t) can be calculated for each time step from the maximum effort capacity values ​​DBAmax(t) for each time step. It can then be presented, for example displayed, to the user.

[0138] Determination method 2 may involve determining the initial maximum exertion capacity DBAi from the first time-domain data S, used to determine the critical intensity DBAc. DBAi is, for example, equal to an absolute record determined from the data S. The data may have undergone a moving average, at least to filter out measurement errors. In such a case, DBAi is, for example, equal to an absolute record for <s>(At —> 0) with the window duration being the smallest considered.

[0139] The fatigue constant T depends on the individual's capabilities. For example, it is equal to 30 s for an adult human or an animal with a body size (e.g., mass) equal to that of an adult human. It can be higher or lower depending on the individual's capabilities or body size. It is generally between 10 s and 1000 s.

[0140] Method 2 can determine the fatigability constant T from the time-domain data S(t) used to determine DBAc. The equation

[0141] Smod(At) = k / (At + T) + DBAc

[0142] with the initial condition

[0143] Smod(At = 0) = DBAl

[0144] The recovery constant R also depends on the individual's capabilities. For example, it is equal to 1.27 for an adult human. It can range from 0 to 10.

[0145] The invention is not limited to the embodiments previously described and extends to all embodiments covered by the invention.< / s> < / ds> < / vxy> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s>

Claims

Demands 1. Data processing method (1), implemented by a server device (3), to obtain a critical effort intensity (DBAc) for an animal, the processing method (1) comprising the steps of: • receive (13), from a client device (4), at least a first time series of data (S) representative of at least one acceleration generated by the animal for several instants (t) over time; • from a plurality of time windows, having distinct durations (At1, At2), calculate (14), for each of the time windows of the plurality of time windows, a convolution product ( <s>(At1), <s>(At2)) of the first time series of data (S) with said time window, called "moving average"; • determine (15) at least one maximum (Smax(At1), Smax(At2)), called the "record", of each of the calculated moving averages ( <s>(At1), <s>(At2)); and • fit (16) a decreasing and convergent model (Smod(At)) on the determined records (Smax), said model being a function of the distinct durations (At1, At2) of the time windows used to calculate the moving averages ( <s>(At1), <s>(At2)); • determine (17) an asymptote (DBAc) of the adjusted (Smod(At)) model, said asymptote (DBAc) being the critical effort intensity (DBAc) of the animal.

2. Processing method (1) according to the preceding claim, wherein the fitting (16) of the asymptotic model (Smod(At)) is carried out by quantile regression with a quantile greater than or equal to 75%.

3. A treatment method (1) according to any one of the preceding claims, further comprising the steps of: • receive a second time series (Z) representing at least one altitude of the animal over time (t), the second time series (Z) being acquired at least partly simultaneously with the first time series (S); • prior to calculating (14) the moving averages, correct (11) the first time series (S) by means of the second time series, each instant of the corrected first time series (S) being a function of the time derivative of the second time series (Z) at the corresponding instant.

4. Processing method (1) according to the preceding claim, wherein the second time series (Z) is representative of at least one barometric data or geolocation data of the animal.

5. Processing method (1) according to one of the two preceding claims, further comprising the steps of: • receiving a third time series representative of at least one geographical data of the animal over time (t), the third time series (XY) being acquired at least in part simultaneously with the first time series (S); • the correction (11) of the first time series (S) being further a function of the third time series (XY), each instant of the first corrected time series (S) also being a function of the time derivative of the third time series (XY) at the corresponding instant.

6. Processing method (1) according to any one of the preceding claims, wherein the first time series (S) is representative of a vector sum of the acceleration generated by the animal as a function of time.

7. Processing method (1) according to any one of the preceding claims, wherein the acceleration generated by the animal as a function of time is sampled at a sampling frequency greater than or equal to 1 Hz.

8. Processing method (1) according to one of the two preceding claims, comprising a step, carried out prior to the calculation of moving averages, of correcting the first time series to remove the effect of gravity suffered by the individual or animal over time.

9. Processing method (1) according to any one of the preceding claims, wherein the duration (At1, At2) of each time window is between five seconds and two hours, and preferably between ten seconds and thirty minutes.

10. Method for determining (2) a critical intensity (DBAc) for an animal, the determination method comprising the steps implemented by a customer device (4): • acquire (21), using an accelerometer, a data point (A) representative of a first acceleration generated by the animal for several instants over time (t), to form a first time series of data (S); and • send (23) the first time series of data (S) to a server device (3) configured to implement the processing method according to one of the preceding claims, so as to obtain the critical effort intensity of the animal, and implemented by the server device (3): • receive (13), from the client device (4), at least the first time series of data (S); • from a plurality of time windows, having distinct durations (At1, At2), calculate (14), for each of the time windows of the plurality of time windows, a convolution product ( <s>(At1), <s>(At2)) of the first time series of data (S) with said time window, called the "moving average"; • determine (15) at least one maximum (Smax(At1), Smax(At2)), called the "record", of each of the calculated moving averages ( <s>(At1), <s>(At2)); and • fit (16) a decreasing and convergent model (Smod(At)) on the determined records (Smax), said model being a function of the distinct durations (At1, At2) of the time windows used to calculate the moving averages ( <s>(At1), <s>(At2)) ;• determine (17) an asymptote (DBAc) of the adjusted (Smod(At)) model, said asymptote (DBAc) being the critical effort intensity (DBAc) of the animal.

11. Method for determining (2) according to the preceding claim, comprising the additional steps, implemented by the customer device (4), of: • acquire (24), using the accelerometer, data relating to a second acceleration generated by the animal during an activity, for several instants over time (t); • compare (26), for at least one instant, the data relating to the second acceleration acquired with the critical effort intensity of the animal so as to obtain a category of the activity; and • indicate (27) the result of the comparison.

12. A method for determining (2) according to the preceding claim, wherein, when the data relating to the second acceleration is less than the critical effort intensity, it indicates that the activity is in a moderate regime; and wherein, when the data relating to the second acceleration is greater than the critical effort intensity, it indicates that the activity is in a vigorous regime. 13.Method of determination (2) according to one of the two preceding claims, wherein the acquisition (24) of the data relating to the second generated acceleration is carried out to form a second time series of data (S2), the method of determination (2) comprising a step of calculation of a maximum effort capacity DBAmax for each time step of the second time series of data (S2) such that: when the second time series of data (S2) at the given time step (t) is greater than or equal to the critical effort intensity (DBAc):. 1 DBAmax(t)=DBAmax(tl)-- (S2(t)-DBAc) T where DBAmax(t-1) is the maximum effort capacity at a previous time step and T is a fatigue constant; or when the second time series of data (S2) at the given time step (t) is less than the critical effort intensity (DBAc) DBAi-DBAmax(tl) DBAmax(t) =DBAmax(tl)+R (DBAc-S2(t)} DBAi-DBAc where DBAi is an initial maximum effort capacity and R is a recovery constant.

14. Server device (3) comprising means for receiving at least a first time series of data (S) from a client device (4), together with a computer configured to carry out the steps of the processing method (1) according to any one of claims 1 to 9.

15. Client device (4) comprising means for sending a first time series of data (S) to a server device (3) together with an accelerometer (41) and a computer (42) configured to implement the steps, relating to the client device (4), of the determination method according to any one of claims 10 to 13.

16. Client-server system comprising a server device according to claim 14 and a client device according to claim 15.

17. Computer program comprising instructions which, when the program is executed by the server device (3) of claim 14, cause the server device (3) to receive at least a first time series of data (S) from a client device (4) and to implement the steps of the processing method (1) according to any one of claims 1 to 9.

18. Computer program comprising instructions which, when the program is executed by a client device (4), cause the client device (4) to send at least a first time series of data (S) to a server device (3) and to implement the steps, relating to the client device (4), of the determination method (2) according to any one of claims 10 to 13.< / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s>