Method and system for automatically generating an individualised sports training schedule

Abstract

The invention relates to a method (500) for automatically generating an individualised electronic sports training schedule (100-1, 100-2) comprising a plurality of sessions for improving the physical performance of a user, the method comprising: a step (510) of collecting individual data of the user via a personal electronic device (200); a step (520) of analysing the collected data by means of machine learning algorithms on a digital analysis platform (300); and a step (540) of generating a first individualised schedule (100-1) on the basis of the collected and analysed data; a step (550) of measuring in real-time, at least during a session, physiological or performance parameters of the user, and / or of modifying the individual data; a step (560) of dynamically adapting the schedule on the basis of the parameters and / or the data of the preceding step by executing machine learning algorithms; and a step (570) of generating a second individualised schedule (100-2) that is better suited to the user.

Description

[0001]Method and system for automatically generating an individualized sports training schedule TECHNICAL FIELD The present invention belongs to the general field of sports technologies, in particular assistance technologies for improving physical performance, and relates more particularly to a method for automatically generating, in real time, an individualized sports training schedule. The generation of such a schedule implements either an expert system (decision tree) or a machine learning model, and is based on various user data. The present invention finds a direct, but not exclusive, application in running training, which is a very widespread discipline among the general public. STATE OF THE ART The practice of a sport, whether in a professional or amateur setting, preferably requires regular training adapted to the objectives sought.Whether it's for intensive competition preparation, improving physical performance, or simply practicing routine exercises, choosing personalized training is particularly important. The promotion of physical activity and its benefits has led to an increase in the number of regular and occasional participants in certain easily accessible disciplines, such as running. For good running practice, particularly in endurance, it is essential to know your physical abilities and follow personalized training to improve your performance. Historically, sports training has undergone significant changes over the 20th century. èmecentury, which continued until the emergence of mobile sports training planning applications. During the 1950s and 1970s, sports training underwent significant changes thanks to advances in medicine and sports science. Coaches began using more accurate methods to measure athletic performance and adapt training accordingly. New training techniques were also developed, such as interval training and block periodization. During the 1980s and 1990s, sports training continued to evolve thanks to advances in technology and sports science. Athletes began using handheld devices to measure their heart rate and fatigue levels, allowing them to better adapt their training. During the 2000s and 2010s, sports training evolved further thanks to the emergence of new technologies.information and communication. Athletes have begun using training tracking apps and connected devices to measure their performance and track their progress. Since 2010, with the development of so-called smart mobile phones (smartphones), mobile sports training planning and tracking apps have emerged. These allow athletes to create personalized training plans based on their goals and fitness level, and track their progress in real time using connected devices. In addition, some apps use a statistical approach to analyze training data and recommend adjustments to the training plan. Nowadays, mobile sports planning apps have become a commonplace solution, increasingly adopted especially by amateur runners. Especially since today's smartphones are carried almost inpermanently and equipped with numerous sensors that allow the wearer to very simply measure various physiological (heart rate, oxygen saturation level, etc.) and kinematic (movement speed, distance traveled, number of steps, cadence, etc.) parameters. These applications are sometimes developed by the phone manufacturers themselves and offer features that go beyond tracking physical activity, such as tracking sleep and other aspects of daily life (nutrition, hydration, stress, etc.). In addition, the integration of artificial intelligence and IoT (Internet of Things) technologies has made it possible to improve sports training applications and create the field of smart sports training (smart fitness). The document "Alireza Farrokhi, Reza Farahbakhsh, Javad Rezazadeh, Roberto Minerva, Application of Internet of Things and artificial intelligence for smart fitness: A survey,Computer Networks, Volume 189, 2021 » presents IoT-based solutions in the field of smart fitness, which are divided into three categories: fitness trackers (including wearable and non-wearable sensors), motion analysis, and fitness applications. According to this document, data collected by users and IoT-based smart fitness devices could be used to improve training performance through artificial intelligence-based algorithms. More specifically, solutions for individualized coaching based on artificial intelligence are known. Document KR102116968 describes a computer-implemented sports training method comprising the following steps: establishing a communication session with the exercise machine via a short-range communication module; receiving the exercise amount information collected in real time fromthe machine; and recommending customized exercise information for the user based on the received exercise amount information and the pre-stored user characteristic information. In this method, the receiving step includes establishing a communication session when the identification information of the user terminal is transmitted to the exercise machine. After receiving the identification information of the exercise machine, while the communication session is maintained, the exercise machine temporarily stores the identification information and the exercise amount information of the corresponding user terminal. When the user performs exercise, the information detected by the sensor of the exercise machines is different for each machine, and the step of recommending the user-specific exercise information is generated based on big data and learningDeep Learning. Document US2021178226 concerns a virtual sports trainer offering dynamic training programs that can make incremental adjustments, based on an athlete's performance, to the training program over time. In addition to the fact that they do not use an expert system or artificial intelligence, these solutions do not allow for certain parameters to be taken into account to adapt the training schedule in real time, in particular the users' menstrual cycle. However, these parameters can have a significant influence on athletic performance. Indeed, the menstrual cycle can have an impact on athletic performance and the body's recovery, so it makes sense to take this factor into account when creating a training schedule. There are a few solutions that integrate the menstrual cycle into the development of a training schedule. For example, the FitrWoman application allowsto track the menstrual cycle and offer personalized training and nutritional suggestions tailored to changes in hormone levels throughout the cycle. For its part, the Wild.AI application takes the menstrual cycle into account by offering a readiness score based on tracking symptoms and well-being, training plans with the option of synchronization with the cycle, individual recommendations on symptoms, pre- and post-workout recommendations, and visualization of data specific to each phase of the cycle. There are also various applications for simple tracking and management of the menstrual cycle, such as MyFLO and Cycle tracking (developed by Apple). We also know the solution of document US2021050086 which describes methods and systems for generating an optimized training plan for an individual based on genetic, physiological, behavioral andlifestyle of the individual. By taking into account the genetic and environmental data of an individual, a training plan can be constructed from individual exercises and exercise parameters in a manner optimized for the physiological traits of the individual. This solution therefore makes it possible to generate training plans based on individual data that are generally fixed (genetic) or variable over the very long term (environment, behavior, etc.), and takes absolutely no account of the physical capacities of the individual at time t, in particular their progress from day to day or from week to week, etc. PRESENTATION OF THE INVENTION The present invention aims to overcome all or part of the drawbacks of the prior art set out above by proposing a solution for individualized planning of sports training, taking into account several physical, physiological and personal factors to adapt the generated schedules in real time. To thisIn effect, the present invention relates to a method for automatically generating an individualized electronic sports training schedule, comprising a plurality of sessions for improving the physical performance of a user, said method comprising: − a step of collecting individual data from the user, via a personal electronic device; − a step of analyzing the data collected by an expert system or by machine learning algorithms, on a digital analysis platform; and − a step of generating a first individualized schedule based on the data collected and analyzed. This method is remarkable in that it comprises: − a step of continuously measuring, at least during a sports training session, physiological or performance parameters of the user, and / or modifying the individual data; − a step of dynamically adapting the schedule based on the parameters and / or data of the stepprevious, by executing the expert system or machine learning algorithms; and − a step of generating a second individualized schedule more suited to the user. According to one aspect of the invention, the individual data collected in the collection step or modified in the modification step include information on the user's menstrual cycle when the user is female. According to one aspect of the invention, the performance parameters measured in the measurement step include information on the user's psycho-physiological state such as the user's level of fatigue or stress. According to one embodiment, the dynamic adaptation step includes a classification or regression of the user's fatigue levels based on cardiac measurements taken by an acquisition device, in particular a connected watch equipped with sensors for this purpose. The cardiac measurements include, for example, measurements ofheart rate variability. According to one aspect of the invention, the dynamic adaptation step comprises a slope break analysis of a signal representing a measured physiological or performance parameter of the user. This parameter is for example the speed during a run. More particularly, the machine learning algorithms executed in the dynamic adaptation step comprise classifiers based on Riemannian geometry, called RGC. The present invention also relates to a system for automatically generating an individualized sports training schedule, for implementing a method as presented, comprising: − a personal communication device, such as a smartphone, for collecting and modifying the individual data of the user, and for displaying the generated schedule; − a data acquisition device, such as a connected watch, for measuring in real time physiological parameters ofthe user; and − a digital analysis and calculation platform for processing the collected and measured data, comprising an analysis engine based on an expert system or a machine learning model for determining a user profile from the collected data; said engine comprising a schedule generator, which may be based on a machine learning model, for generating an individualized training schedule using the user profile and historical training data, and an adaptation module for dynamically adapting the training schedule based on the current training data and subjective feedback from the user to continue to individualize the schedule. Thus, the present invention makes it possible to collect data on physical characteristics, training objectives, availability, unforeseen events, subjective feedback, successes, failures, overall performance, fatigue,recovery, sleep, nutrition, interactions between different sports disciplines and information on the user's menstrual cycle, this can be manual or automatic via physiological measurements; to analyze the collected data using artificial intelligence to determine an individualized training profile adapted to the user's menstrual cycle; to generate an individualized training schedule using a machine learning model based on the individualized training profile; and to dynamically adapt the training schedule based on current training data, user feedback, physiological data acquired via portable or remote sensing devices and information on the user's menstrual cycle to continue to individualize the schedule by adapting the intensity, duration, type of training, terrain, exercises, etc. The method canalso include a step of monitoring the progress of the training and sending notifications and advice to the user, in particular for educational reasons. The fundamental concepts of the invention having just been set out above in their most basic form, other details and characteristics will emerge more clearly on reading the description which follows and with reference to the appended drawings, giving by way of non-limiting example an embodiment of a method for automatically generating an individualized sports training schedule, in accordance with the principles of the invention. BRIEF DESCRIPTION OF THE FIGURES The figures are given purely for illustrative purposes for a better understanding of the invention without limiting its scope. The various elements may be represented schematically and are not necessarily to scale. Throughout the figures, identical or equivalent elements bear the same numerical reference. Itis thus illustrated in: − Figure 1: a flowchart of the main steps of a method for automatically generating an individualized sports training schedule according to one embodiment of the invention; − Figure 2: a diagram of the implementation of the method according to a simplified scenario; − Figure 3: a diagram of the factors for individualizing the training schedule according to the invention; − Figure 4: an overall architecture of an information and communication system for implementing the method according to the invention; − Figure 5: an example of a graphical interface of a mobile application dedicated to the method; − Figure 6: another example of a graphical interface of a web page dedicated to the method. DETAILED DESCRIPTION OF EMBODIMENTS It should be noted that certain technical elements well known to those skilled in the art are described here to avoid any insufficiency or ambiguity in understanding the present invention. In the embodiment described below,refers to a method for automatically generating, on a personal electronic device such as a smartphone, an individualized sports training schedule, intended primarily for the practice of running. This non-limiting example is given for a better understanding of the invention and does not exclude the use of the method in the practice of any other sporting discipline or activity requiring regular training. In the present description, and unless otherwise indicated, the following definitions are adopted: − A “model” designates an abstract concept based on data and having hyperparameters, which is characterized by an output of interest obtained from input data. A typical example of a model is a machine learning model. − An “instruction” is a series of commands verifying the following properties: i. the commands are grouped on the basis of a function; ii. the series of commands is a component of acomputer program; and iii. the series of commands is executable by a processor. All other technical and scientific terms used have the same meanings as those commonly accepted in the technical field of the invention. Also, “collection”, “analysis”, “elaboration”, “generation”, “adaptation”, “calculation”, and their derived forms, or more broadly “executable operation” within the meaning of the invention, means an action performed by a device or processor unless the context indicates otherwise. In this regard, operations refer to actions and / or processes of a data processing system, for example a computer system or an electronic computing device, which manipulates and transforms data represented as physical (electronic) quantities in the memories of the computer system or other devices for storing, transmitting or displaying information. These operations may bebased on applications or software. To facilitate the reading of this description in relation to the figures, the main steps of the method which is the subject of the present invention will be presented in their general form with reference to Figure 1. The implementation of the method according to a simplified scenario is then illustrated in Figure 2. Examples of individualization factors are given in Figure 3. The overall architecture of an information system allowing the implementation of the method will then be presented with reference to Figure 4. Finally, possible graphical interfaces will be presented with reference to Figures 5 and 6. Figure 1 in fact represents the main steps of a method 500 for automatically generating an individualized sports training schedule, said method comprising: − an initial step 510 of collecting individual data from the user, via a personal electronic device 200; − a step 520 of analyzing the datacollected by a machine learning model, on a remote or embedded digital analysis platform 300; − a step 530 of developing a user profile; − a step 540 of generating a first individualized schedule 100-1 based on the data collected and analyzed; − a step 550 of measuring in real time, at least during a sports training session, physiological parameters or performance of the user, and / or modification of the individual data; − a step 560 of dynamic adaptation of the schedule based on the parameters and / or data of the previous step; and − a step 570 of generating a second individualized schedule 100-2 more suited to the user. Of course, steps 550 to 570 are iterative and can be repeated several times until the end of the training targeted by the user. Step 510 of collecting individual user data consists of collecting information about thecharacteristics and preferences of the user, such as his gender, age, weight, physical condition, sporting objectives, etc., using a personal electronic device 200, typically a smartphone on which an application dedicated to the implementation of the method is installed. The individual data, which may be collected initially, include, but are not limited to: − information on the identity of the user (name, contact details, date of birth, city or country of residence, etc.); − information on the user's physique (weight, height, etc.); − information on the sporting objectives targeted by the user (discipline, deadline, etc.); − information on the sporting habits of the user (sports practiced, frequency of training, daily or weekly duration of physical activity, etc.); − information on the user's daily lifestyle habits (diet, sleep, time slotsavailability for sports training, etc.); − medical and / or general health information of the user and / or all or part of the physical and physiological data collected during past sports sessions when said data is available; − information on the menstrual cycle of users; − etc. Of course, each user has the choice of providing or not providing any information that he or she deems unnecessary, except for non-sensitive information necessary for the implementation of the method, in particular purely sporting information such as the chosen discipline, the objective and the time slots of availability. The information collected is then analyzed and used to develop an evolving user profile and generate a first individualized sports schedule. This is at this stage a first level of individualization, advantageously enabled by the present invention, insofar as all theThe individual data initially collected are analyzed and processed in a machine learning model to obtain the most representative user profile possible and to enable the generation of the most suitable schedule possible. Step 520 of analyzing the collected data consists of using one or more machine learning models to analyze all or part of the data collected in step 510. This makes it possible to understand the characteristics and preferences of the user in a more in-depth manner and to use them to develop a more precise user profile. This analysis is carried out on a digital analysis platform 300 which may be remote, in particular in the cloud, or embedded on the personal electronic device 200 which makes it possible to execute the application associated with the method. During step 520, the machine learning model uses specific algorithms to analyze the data collected on the user, such asas information about the user's habits and health. These algorithms may include techniques such as regression, principal component analysis (PCA), k-nearest neighbor analysis (k-NN), etc. The analysis makes it possible to understand the relationships between the different data and use them to develop a more accurate user profile. The profile is then used to generate an initial individualized training plan for the user. Step 530 of developing a user profile consists of creating a summary of the user's characteristics and preferences using the information collected and analyzed in steps 510 and 520. The user profile may include information such as the user's activity level, athletic goals, hours of availability, injury history, etc., presented in a standardized manner on the dedicated application. For example, for a 30-year-old user weighing 75kg, who is a beginner runner and has a goal of running a half marathon in 6 months, the application analyzes the collected data and develops a profile that describes the user as a beginner runner with a medium-term goal of running a half marathon. This user profile is then used to generate a first individualized training plan for the user. Step 540 of generating a first individualized plan 100-1 consists of using the user profile developed in step 530 to generate a personalized sports training plan for the user. This can be achieved using planning algorithms such as linear programming algorithms, genetic algorithms or operations research algorithms. During this step, the algorithms can take into account information such as the user's activity level, his sports goals, his hours of availability, his history ofinjury, the menstrual cycle of female users, etc. They can also take into account criteria such as training load or volume, training frequency, training duration, etc. The final objective is to generate a personalized sports training plan that best suits the needs and objectives of the user, while taking into account their physical limitations and specificities and their availability schedules. This first schedule can be displayed to the user via the dedicated mobile application, on a web interface or on any other suitable connected object (smartwatch for example). This is at this stage a second level of individualization, advantageously enabled by the present invention, insofar as the generation of the first schedule takes into account the basic level, availability, unavailability due to unforeseen events, eating and sleeping habits of the user. Step 550 of measurement inreal-time monitoring consists of monitoring the user's physiological and performance parameters during a sports training session programmed in the first individualized schedule. This can be achieved by using multi-sensor acquisition devices 400, including for example heart rate sensors, motion sensors, sweat sensors, pedometers, etc. The acquisition device 400 can be a connected watch, a chest strap, a connected ring, a connected bracelet, or any other patch worn on the user's skin. These personal, so-called "smart" devices use different technologies to measure various physiological parameters. When it comes to parameters measured in or by the blood such as heart rate, oxygen saturation, blood glucose level, lactate level, etc., the acquisition device 400 can use non-invasive technologies, includingphotoplethysmographic (PPG), or certified invasive devices such as patches equipped with micro-needles. These devices send the measured data in real time or delayed to the dedicated application, which analyzes them to evaluate the user's performance during the training session. The collected data can also be used to update the user's individual information, such as injury history, sports goals, etc. The objective of this step is to enable real-time evaluation of the user's performance to adjust the training plan accordingly. This can help ensure that the user achieves their sports goals more efficiently and safely. In step 550, the user can also modify certain pre-recorded parameters and update their personal data (goals, availability, etc.). Of course, manual user intervention formodifying one's data remains possible at any time. When measuring parameters, the intermediate information deduced is processed in raw form on the computing platform 300, but also pre-processed (in terms of filtering, analysis, derivation of indicators) directly in the personal device 200 or by other remote software. This intermediate information can be used as indicators of performance, feelings, fatigue, health status, stress status, sleep quality, or nutrition quality. The step 560 of dynamic adaptation of the schedule is the logical continuation of the step 550 of real-time measurement, and uses the data collected on the performance and physiological parameters of the user during the training session to adapt the training schedule in real time. This can be done by using machine learning algorithms to adjust the training sessions according to the datacollected. For example, if the user seems tired during a session, the application may decide to adjust the exercises for the next session to take this factor into account. The objective of step 560 is to enable the generation of a more personalized training plan that adapts in real time to the user's abilities and objectives. This can help maximize results and minimize injuries. This is at this stage a third and final level of individualization, advantageously enabled by the invention, insofar as the adaptation step takes into account: − the user's subjective feedback (for example "this workout was too hard, too short, not suitable, etc."); − the user's progress by periodically recalculating the schedule based on successes, failures, and overall performance in past training sessions; − the user's fatigue and recovery; − the interactions between thedifferent sessions of different disciplines (cycling, running, swimming, etc.). Dynamic adaptation refers to the continuous and real-time adjustment, in the analysis platform, of the training schedule based on the data collected on the user. This means that the platform constantly analyzes information about the user, such as physiological data, performance, subjective feedback, unforeseen events, successes, failures, fatigue, recovery, sleep, nutrition, interactions between different sports disciplines and information on the menstrual cycle to ensure that the training schedule is adapted to the user's needs in real time. Adaptation is considered dynamic because it is done continuously, in real time and automatically. In other words, the schedule does not require direct human intervention to adjust, it is constantly adapting to the collected data.This allows the training results to be optimized based on the user's progress and needs. The step 570 of generating a second individualized schedule 100-2 is the final step of the method described. It uses the information collected and analyzed in the previous steps to generate a fully personalized training plan for the user. This can be done using machine learning algorithms that take into account all the data collected on the user's training habits, performance, and physiological parameters. The objective is to provide a schedule that takes into account the user's abilities and goals to achieve the best possible results. Different algorithms can be used, including predictive algorithms such as the ordinary least squares method, regularization methods, the Lasso method, logistic regression,decision tree forest, gradient boosting, a support vector machine, a stochastic gradient algorithm, a k-nearest neighbor method; and data classification and partitioning algorithms such as K-means, the Gaussian mixture model, spectral partitioning, the DBSCAN algorithm, interactive partitioning, an isolation forest for the detection of physiological anomalies during training sessions, etc. Figure 2 shows the implementation of the method according to a simplified scenario in order to understand the principle. Initially, on a date T0, a first individualized schedule 100-1 is generated from individual data provided by a user U who can enter them on a smartphone 200 for example. The first schedule 100-1 contains at least one training session S scheduled on date T1, with a duration ^T. The completion of this first session by the user makes it possible to measure dataphysiological or performance data using a connected watch 400 for example. This makes it possible to update the schedule in real time and therefore generate a second schedule 100-2, and so on until the user reaches their training objective. As a result, the schedule updates, which correspond to the dynamic adaptation step repeated as many times as sessions performed and manual modifications to the profile, make it possible to obtain multi-factor individualization of the sports training schedule. Figure 3 shows the different factors for individualizing sports schedules generated by the method according to the invention. Some factors, called static, correspond to the initial data collection step 510 and include: − the base level 511; − sleep and nutrition 512; − The availability and menstrual cycle of the users 513. These first factors can be entered using an electronic devicesuch as a smartphone. Other factors, called dynamic, correspond to the steps 550 of dynamic adaptation of the schedule and include: − fatigue and recovery 551; − subjective feedback 552; − interactions between the different sports practiced 553; − unforeseen events 554; − progression 555. These second factors can be informed, depending on their type, either automatically by acquisition devices such as a connected watch, or manually by personal electronic devices such as a smartphone. Of course, the “static” factors become “dynamic” when they evolve over time, as the user’s training progresses. Regarding fatigue 551 as an individualization factor, the method makes it possible to follow the evolution of the user’s fatigue according to the training loads and to adapt the training to the level of fatigue according to learning modelsautomatic. These models allow the classification of fatigue states via cardiac measurements, in particular electrocardiogram (ECG) or plethysmography (PPG) measurements taken during training sessions. The level of fatigue can be estimated using physiological parameters including heart rate variability, exercise-related cardiac dynamics and exercise-related heart rate, but also using non-physiological parameters such as performance indicators. The classification of fatigue states by cardiac measurements can be carried out by an algorithm based on Riemannian geometry, preferably a Tangent Space Classifier (TSC). Riemannian approaches represent ECG measurements as covariance matrices, which are symmetric positive definite (SPD) matrices, and manipulate them with an appropriate geometry, the geometryRiemannian. Classifiers based on such geometry are called Riemannian Geometry-based Classifiers (RGC). First, in a Riemannian manifold, the intrinsic non-Euclidean distances between two SPD matrices, i.e., two points (here C1 and C2), can be estimated using the Riemannian distance: In the above formula, ^ ^ ^^^ is the nth eigenvalue of the matrix M. The set of tangent vectors at point G on the manifold defines the tangent space to the manifold at G. More generally, any SPD matrix Ci can be projected onto the tangent space at point G using: Si being the projection of Ci on the tangent plane, and logm denoting the logarithm of a matrix. The cardiac signal is thus represented in the form of a covariance matrix, the projection of which in a Euclidean space makes it possible to classify the levels of fatigue (or others) by measuring distances in this space. On the other hand, concerning the menstrual cycle 513, the training schedule is adapted thanks to the measurement and knowledge of the cycle, in particular by the manual information provided by women, by automatically analyzing physiological measurements during the cycle.This adaptation of training to the cycle is based in particular on: − the recovery of subjective information on the feelings and / or capacities of women, compared to the phases of the cycle; − the recovery of objective information on changes in performance, compared to the phases of the cycle; − the recovery of physiological information on performance and physiological response to exercise in women, compared to the phases of the cycle; − the direct optimization of training based on prior scientific knowledge on the best training depending on the phases of the cycle (optimization can be understood in terms of performance during training, in terms of optimal physiological progress thanks to these training sessions, or in terms of feelings during training).Furthermore, the concept of adaptation of the sports training schedule encompasses the following aspects: − regarding what is meant by "adaptation of a training": adaptation of intensity, length (volume), type of training, terrain, exercises, time spent in each intensity zone, etc.; − regarding what can be adapted: a training phase, a training session, a training week, a cycle (i.e. several weeks) of training; − regarding the time of adaptation: adaptation week by week, day by day, minute by minute, etc.; − regarding the type of adaptation: the adaptation can be an imposed, proposed, suggested modification, or the proposal of one or more training alternatives; − regarding what accompanies this adaptation: it can be explained to the user so that he makes a definitive choice by completing this proposal with a feeling.Figure 4 represents the overall architecture of a system for communicating and exploiting user U data, comprising a personal communication device 200, a computing platform 300 and an acquisition device 400. This system is organized into three paradigms: data acquisition, data exploitation and information presentation. The parameters measured by the acquisition device 400 are sent, in real time or at regular time intervals, to the analysis platform 300 by means of a wireless network. At the request of the computing means of said platform, parameters are transmitted to them by servers. These means then carry out series of calculations by executing dedicated programs.Among the programs installed on the platform 300, some perform a simple presentation, interactive or not, of the collected data, others perform a multifaceted interpretation and representation of the data, and still others compile artificial intelligence algorithms to predict information not available in the state. It should be noted that machine learning generally requires a large amount of data to function effectively. Thus, to use machine learning in a method of automatically generating individualized sports training plans, it will probably be necessary to collect and store large amounts of data on the characteristics of the user and on his training, as well as data from a community of users.In summary, according to the invention, machine learning is used to predict the user's athletic performance based on their physical characteristics and training history. This prediction can be used to adapt the user's training schedule and to give them realistic goals. Machine learning is also used to optimize the user's training based on their physical characteristics and athletic goals. For example, a machine learning model can be trained on body performance and recovery data to recommend optimal training intensity and volume levels for the user. In addition, machine learning is used to adapt the user's training in real time, based on their training reactions and fatigue levels.For example, a machine learning model can be trained on heart rate and fatigue level data to recommend adjustments to the training schedule in real time. Figure 5 shows an example of a dashboard 210 for a mobile application, displayed on a smartphone 200. In this reduced display case, the dashboard preferably displays a few sections per window for better readability. Here the weekly section 211 is displayed with different information modules 212 to 214. Quick access keys 215 can also be provided to show relevant data directly linked, for example, to sports training sessions. Those skilled in the art will easily understand that the dashboard can be adapted according to the needs of each user. Figure 6 shows another example of a user interface 250, more suited to a web display, the ergonomics of which allow simplified and intuitive navigation.It is clear from this description that certain non-essential elements of the method and its implementation system may be modified, replaced or deleted, and that certain steps of the method may be adjusted to other use cases without departing from the scope of the invention defined by the claims below.

Claims

CLAIMS ________________ 1. Method (500) for automatically generating an individualized electronic sports training schedule (100-1, 100-2), comprising a plurality of sessions for improving the physical performance of a user, said method comprising: − a step (510) of collecting individual data from the user, via a personal electronic device (200); − a step (520) of analyzing the collected data by an expert system or by machine learning algorithms, on a digital analysis platform (300); and − a step (540) of generating a first individualized schedule (100-1) based on the collected and analyzed data; characterized in that it comprises: − a step (550) of continuously measuring, at least during a sports training session, physiological or performance parameters of the user, and / or modifying the individual data;− a step (560) of dynamically adapting the schedule according to the parameters and / or data of the previous step, by executing the expert system or machine learning algorithms; and − a step (570) of generating a second individualized schedule (100-2) more suited to the user.

2. Method according to claim 1, wherein the individual data collected in the collection step (510) or modified in the modification step (550) comprise information on the menstrual cycle of the user when the user is female.

3. Method according to claim 1 or 2, wherein the performance parameters measured in the measurement step (550) comprise information on a physiological state of the user such as the level of fatigue or stress.; 4. Method according to any one of the preceding claims, wherein the dynamic adaptation step (560) comprises a classification of fatigue levels of the user according to cardiac measurements carried out by an acquisition device (400).

5. Method according to claim 4, wherein the cardiac measurements comprise cardiac variability measurements.

6. Method according to any one of the preceding claims, wherein the dynamic adaptation step (560) comprises a slope break analysis of a signal representing a measured physiological or performance parameter of the user.

7. Method according to any one of the preceding claims, wherein the machine learning algorithms executed in the dynamic adaptation step (560) comprise classifiers based on Riemannian geometry, called RGC. 8.System for automatically generating an individualized sports training schedule (100), for implementing a method according to one of the preceding claims, comprising: − a personal communication device (200) for collecting and modifying the individual data of the user, and displaying the generated schedule; − a data acquisition device (400) for measuring physiological parameters of the user in real time; and − an analysis and calculation platform (300) for processing the collected and measured data, comprising an analysis engine based on an expert system or a machine learning model for determining a user profile from the collected data; said engine comprising a schedule generator for generating an individualized training schedule using the user profile and. historical training data, and an adaptation module to dynamically adapt the training schedule based on current training data and subjective user feedback to further individualize the schedule.

Images