System for phenotyping-based sports training

Abstract

The invention relates to a method for the automated optimization of sports training, wherein a portable non-invasive high-density electromyography sensor unit (1) is attached to a person (S1), the person performs a sports movement sequence to be optimized while wearing the sensor unit (1) (S2), data is recorded by the sensor unit (1) during the target movement sequence (S3), the data is evaluated by a computing unit (2) for information about active motor neurons (S4) and stored in a personal database (S5); and wherein, based on this, the computing unit (2) determines a training movement sequence (S6) that is different from the target movement sequence, but exhibits a minimal difference between the training movement sequence and the target movement sequence with respect to active motor neurons or muscle fibers activated by the motor neurons, according to a pre-stored muscle-building model.

Description

[0001] The invention relates to a method for the automated optimization of a sports training program, as well as a system for the automated optimization of a sports training program.

[0002] An athlete's training is typically heavily based on their coach's experience. Specific muscle groups are often trained through targeted exercises like deadlifts or similar movements, with the expectation of increased strength in the involved muscle groups. However, such exercises are rather unspecific to the athlete's final movement, yet they engage similar muscles as the primary muscle groups involved in the final movement being optimized. For example, a sprinter doesn't train their thigh muscles solely through sprinting, but also through weight training.

[0003] Based on biomechanical analyses of muscle strength, particularly intramuscular coordination according to recruitment, frequency, and synchronization, both the athlete's coordinative technique (i.e., synchronization) can be improved, and the current state of the musculature (i.e., recruitment and frequency) can be fundamentally examined. Imaging sensors such as ultrasound, as well as other myographic sensors of electromyography, can be used for this purpose.

[0004] Non-invasive wearable electromyography is described, for example, in the white paper of the Meta Group, published on January 9, 2025: “Human-Computer Input via a Wrist-Based sEMG Wearable,” https: / / www.meta.com / dede / blog / quest / surface-emg-wrist-white-paper-reality-labs / .

[0005] See also the publication: A generic noninvasive neuromotor interface for human-computer interaction CTRL-labs at Reality Labs, David Sussillo, Patrick Kaifosh, Thomas Reardon bioRxiv 2024.02.23.581779; doi: https: / / doi.org / 10.1101 / 2024.02.23.581779

[0006] In the field of electromyography (EMG), it is also known to analyze muscle fiber levels and apply neuronal decomposition algorithms that can deduce the activation of individual motor neurons from the sum signal of motor units, representing the lowest level of muscle dynamics. This allows fast-twitch fibers, in particular, to be reliably and repeatedly identified from the sum signal. This is explained, among other things, in the following publications: - Larger and denser: an optimal design for surface grids of EMG electrodes to identify greater and more representative samples of motor units Arnault H. Caillet, Simon Avrillon, Aritra Kundu, Tianyi Yu, Andrew T.M. Phillips, Luca Modenese, Dario Farina bioRxiv 2023.02.18.529050; doi: https: / / doi.org / 10.1101 / 2023.02.18.529050 - A convolutional neural network to identify motor units from high-density surface electromyography signals in real time. Yue Wen, Simon Avrillon, Julio C Hernandez-Pavon, Sangjoon J Kim, François Hug and José L Pons. Published 6 April 2021. Journal of Neural Engineering, Volume 18, Number 5. Yue Wen et al 2021 J. Neural Eng. 18 056003 DOI 10.1088 / 1741-2552 / abeead - Holobar, Aleš and Damjan Zazula. „Multichannel Blind Source Separation Using Convolution Kernel Compensation.“ IEEE Transactions on Signal Processing 55 (2007): 4487-4496.

[0007] From US 2021 / 0315509 A1, a decomposition of a sum signal is also known and relates to a device comprising: a neuro-interface for retrieving surface electromyography signals of a nervous system; a training module that generates a separation matrix based on the first electromyography signals obtained over a first period; and a decomposition module for detecting one or more motor neuron action potentials for individual motor neurons based on second electromyography signals and the separation matrix, wherein the second electromyography signals are generated over a second period that is shorter than the first period, and produce an output in the form of a time series that indicates the activity of the motor neuron.

[0008] Furthermore, solutions already exist for embedding high-density EMG sensors in wearable tissue, primarily to facilitate and improve the reading of electromyographic signals in rehabilitation applications. See the following publications for more information: - Identifying alterations in hand movement coordination from chronic stroke survivors using a wearable high-density EMG sleeve. Nicholas Tacca, Lan Baumgart, Bryan Schlink, Ashwini Kamath, Collin Dunlap, Michael Darrow, Samuel Colachis IV, Philip Putnam, Joshua Branch, Lauren Wengerd, David A Friedenberg, Eric C Meyers. medRxiv 2024.01.02.24300714; doi: https: / / doi.org / 10.1101 / 2024.01.02.24300714. - Gea Drost, Dick F. Stegeman, Baziel GM van Engelen, Machiel J. Zwarts, Clinical applications of high-density surface EMG: A systematic review, Journal of Electromyography and Kinesiology, Volume 16, Issue 6, 2006, Pages 586-602, ISSN 1050-6411, https: / / doi.org / 10.1016 / j.jelekin.2006.09.005. - D. Farina, T. Lorrain, F. Negro and N. Jiang, „High-density EMG E-Textile systems for the control of active prostheses,“ 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology, Buenos Aires, Argentina, 2010, pp. 3591-3593, doi: 10.1109 / IEMBS.2010.5627455.

[0009] DE 10 2019 120 655 A1 further relates to a method for generating individual training sequences in a simulation environment, particularly for sports and rehabilitation purposes, wherein the method comprises the following steps: acquiring user data via user input and sensors; analyzing the user data and determining calculated body and / or movement data; selecting training data corresponding to the calculated body and movement data from a first data store, whereby the physiological effect on the user is a selection criterion; and generating a training specification; providing exercise modules in a second data store; combining and arranging the selected exercise modules according to the specifications of the training specification; generating a virtual environment, environment visualization and 3D modeling from a simulation logic selected by the user; and starting the training.The procedure further comprises the following steps: a) capturing user information, such as body data and movement parameters, during training; b) comparing the captured user information with the calculated data; c) adapting the training specification to the target data; d) adjusting the combination and arrangement of the selected exercise modules according to the specifications of the adapted training specification; e) adapting the virtual environment to the modified exercise modules; f) executing the adapted combination and arrangement of the selected exercise modules and jumping to step a).

[0010] DE 10 2014 118 434 B4 relates to a device for monitoring bodily performance, comprising: at least one communication module with a device for providing an output direct current by means of a direct current line, and a device for receiving sensor signals representative of a measured physiological activity from at least one sensor unit; at least one sensor unit configured for measuring a physiological activity and for generating analog measurement signals responding thereto, wherein the sensor unit is connected to the direct current line and has a device for processing and transmitting sensor signals representative of the measured physiological activity via a line to the communication module; a processing unit configured to form data messages from the analog measurement signals, which are assigned to successive predetermined transmission time slots of a wireless communication protocol;a wireless communication unit for transmitting data messages in the transmission time slots in a chronological sequence based on the wireless communication protocol; and a wireless receiving device configured to receive data messages in successive time slots.

[0011] EP 3 069 656 B1 relates to a system for recording and analyzing the muscle activity of an individual, comprising at least one electromyographic recording section capable of recording, by means of respective sensors, at least initial electrical signals of the muscle group of an individual; a garment connected to the recording section, equipped with electrodes and an electromyographic surface recording card (EMG) with one or more channels for each electrode or pairs of electrodes capable of recording, filtering, and processing the electrical signal emitted during muscle contraction; a computer processor further comprising analytical and processing means for analyzing and processing the recorded signals; a user interface through which output processed by the computer processor can be provided; and a communication interface with the said electromyographic section.and wherein at least one sensor for detecting the impedance of the individual's skin is further provided to determine an impedance value; wherein the computer processor further comprises a database of deductive rules; the processing and analysis means are equipped with an expert system in the form of a computer system capable of reproducing the performance of one or more persons experienced in a particular field of activity, comprising an inference engine by which deductive rules, appropriately stored in the database, are applied to the data obtained from the analysis and processing means, and by which the system offers to use such an impedance value as a correction to the gain of the sensors of the electromyographic recording section.

[0012] WO 2016 / 131936 A2 concerns a device for receiving, in particular reading, electromyography signals and / or transmitting electrical muscle stimulation signals to a human body for training purposes.

[0013] US Patent 2021 / 0283001A1 relates to devices for the rehabilitation of the musculoskeletal system, e.g., the limbs. In particular, it discloses a device, specifically a robotic platform, capable of optimizing gravity-dependent trunk movements, enabling locomotion above the ground for non-ambulatory individuals with spinal cord injury and stroke, and simultaneously promoting sustained motor improvement when mediated during gait rehabilitation facilitated by electrical spinal cord stimulation.

[0014] US 2021 / 0145 302 A1 concerns an electromyography (EMG) device comprising a mainboard with opposing first and second faces. A plurality of first connectors of a first type are provided on the first face, and a plurality of input contacts are provided on the second face. An EMG circuit is located on the mainboard. The EMG circuit is configured to use the input contacts as inputs to receive an EMG input signal and to process the EMG input signal to provide an EMG output signal that is based on, but different from, the EMG input signal. For each of the input contacts, there is no direct conductor path between the input contact and any of the first connectors.

[0015] Each muscle fiber is associated with a motor neuron and forms a functional unit with it. When a motor neuron is activated, a signal is transmitted from the motor neuron to the muscle fibers coupled to it. There, through a cross-bridge cycle, ATP energy is converted into muscle force. Typically, different fiber types are distinguished: slow-twitch and endurance fibers, and fast-twitch and power fibers.

[0016] The maximum force a motor unit can generate therefore depends on the connected muscle fibers, all of which are activated simultaneously. The human body has muscles with relatively few motor units, such as the finger muscles. These do not allow for fine gradations in the force that can be achieved. In contrast, there are muscles with a very large number of motor units, such as the biceps brachii with around 800 motor units.

[0017] The object of the invention is to create and, in particular, to be able to monitor a training movement sequence for an athlete that is optimally tailored to the use of the muscle fibers.

[0018] The invention is defined by the features of the independent claims. Advantageous further developments and embodiments are the subject of the dependent claims.

[0019] A first aspect of the invention relates to a method for the automated optimization of sports training, wherein a portable non-invasive high-density electromyography sensor unit is attached to a person, the person performs a sports movement sequence to be optimized while wearing the sensor unit, data is recorded by the sensor unit during the movement sequence, the data from the sensor unit is evaluated by a computing unit to obtain information about active motor neurons or about muscle fibers of the person activated by the motor neurons during the movement sequence, and the information is stored in a personal database;and wherein the computing unit determines a training movement sequence for the person based on the stored information, which is different from the target movement sequence, but according to a pre-stored muscle-building model exhibits a minimal difference with respect to active motor neurons or muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence, wherein the automatic determination by the computing unit is carried out in such a way that the training movement sequence is maximized to train precisely those muscle fibers which are explicitly used in the target movement sequence according to the database, or are implicitly considered activated by means of the identified predominantly active motor neurons.

[0020] The wearable, non-invasive high-density electromyography (H-EMG) sensor unit is preferably designed as a "wearable" device, such as a sock, sleeve, trouser leg, glove, or similar item. Electrodes placed on or within tissue allow for the application of electromyography (EMG) to electrophysiologically evaluate muscle activity. Non-invasive means that the sensor unit is positioned superficially, eliminating the need to insert electrodes into the tissue; therefore, this type of sensor is also called "surface electromyography."

[0021] In a further embodiment, the sensor unit is controlled by the computing unit in such a way that either all or only a true subset of all electrodes of the sensor unit are used to determine muscle activity.

[0022] High-density electromyography (HEM) also makes it possible to differentiate muscle activity according to active motor neurons. Since each motor neuron controls a multitude of specific muscle fibers, it is also possible to identify the muscle fibers used during an athlete's movement—a process known as phenotyping. Different fibers have characteristic properties; for example, some are fast and powerful, while others are slow and endurance-oriented. HEM allows for the differentiated examination of their activation during a person's movement over time.

[0023] It is assumed that a person uses athletic training to improve a specific athletic movement sequence. Such a sequence could be, for example, the pedaling motion of a cyclist, the start or stationary running motion of a sprinter, the jumping motion for the long jump, a movement in swimming, shot put, high jump, javelin throw, or similar activities. The target movement sequence is thus the movement sequence typically performed in an athletic competition, and the goal is to optimize it. In this context, optimization means performing the target movement sequence in such a way that it leads to better times, greater distances, higher speeds, improved execution of gymnastic movements, or similar target parameters.

[0024] However, a person will not limit their athletic training solely to the target movement sequence to be optimized; rather, a training sequence is used to adapt the person's physiology in such a way that the targeted execution of the training sequence ultimately optimizes the target movement sequence. For example, as mentioned earlier, the speed at the start of a sprint from starting blocks, in terms of the target movement sequence, can be improved by deadlifts, in terms of the training sequence.

[0025] The procedure is thus divided into two phases: First, the high-density electromyography (HEM) sensor unit is attached to the individual while they perform the actual athletic movement sequence to be optimized. During the execution of the movement sequence, the sensor unit records data on the activation of individual muscle fibers. The data is then analyzed by the processing unit and stored in a personalized database, allowing it to be reassigned to that individual when they later perform a similar movement sequence. The information stored in the database indicates which motor neurons and / or which muscle fibers activated by these motor neurons are primarily active during the individual's movement sequence. A threshold value can be used to categorize the muscle fibers as "primarily active" or "less active to irrelevant" with regard to their activation level.

[0026] The database thus reflects muscular activity that can be assigned to the athletic target movement sequence to be optimized, such as a runner's sprint. In a further phase of the process, the processing unit assigns the corresponding data set to a person currently training and automatically determines a training movement sequence that differs from the target movement sequence. For example, the target movement sequence is a sprinter's start from the starting blocks, while the training movement sequence is performed in a gym using equipment, such as a leg press or a barbell for deadlifts.The automatic determination by the computing unit takes place in such a way that the training movement sequence is maximized to train precisely those muscle fibers which are explicitly used in the target movement sequence according to the database, or which are implicitly considered activated by means of the identified mainly active motor neurons.

[0027] It is an advantageous effect of the invention that an optimization algorithm supports a coach, or optionally the athlete himself, in finding an optimal training movement sequence in order to prepare a target movement sequence to be optimized for a competition during training.

[0028] Furthermore, if the wearable, non-invasive high-density electromyography (HEM) sensor unit is worn during the execution of the training movement sequence, the processing unit can provide real-time feedback on whether the movement is being performed ideally to activate the targeted motor neurons. This feedback is preferably displayed on a screen, but can also be provided audibly or haptically. Vibration motors can be attached to a wearable piece of fabric containing the sensor unit and activated depending on the direction of movement, position, or location of the body part being trained, depending on the need for correction.

[0029] According to an advantageous embodiment, the computing unit selects the training movement sequence from a multitude of pre-stored possible training movement sequences that exhibits a minimal difference with respect to the active motor neurons or the muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence.

[0030] According to a further advantageous embodiment, the training movement sequence is assembled by the computing unit by a combination of pre-stored training movement sequence elements in such a way that the resulting training movement sequence has a minimal difference with respect to the active motor neurons or the muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence.

[0031] A training movement sequence generally comprises at least two training movement sequence elements, but typically a significantly higher number. Furthermore, the processing unit preferentially parameterizes the training movement sequence elements in such a way as to minimize the difference. Parameterization of the assembled training movement sequence elements includes, in particular, a duration, an intensity, and / or the movement pattern of the person during the execution of each training movement sequence element. Thus, the combination of training movement sequence elements and their parameterizations provides a multitude of degrees of freedom, which are optimization variables for the processing unit to minimize the difference as explained above.

[0032] According to a further advantageous embodiment, the training movement sequence is generated by the computing unit by applying a generative pre-trained machine learning model.

[0033] The generative pre-trained machine learning model is preferably based on a transformer architecture, although other architectures are also applicable. An artificial neural network is preferably used to generate an optimized training movement sequence, minimizing the aforementioned difference as much as possible. Thus, the problem of generating the training movement sequence for the computing unit no longer consists of a combinatorial selection of predefined elements, which may be parameterizable, but rather a generative process is applied, in which the computing unit can create training movement sequences that are either restricted or more freely selectable. It is also possible that the computing unit, using the generative pre-trained model, will determine a training movement sequence that is entirely new and was not present in any training dataset during its pre-training.

[0034] According to another advantageous embodiment, the computing unit adapts the machine learning model during the execution of the training movement sequence to minimize the difference in active motor neurons or muscle fibers activated by motor neurons between the training movement sequence and the target movement sequence. This corresponds to the approach of reinforcement learning. This allows the machine learning model to be further optimized using information acquired during the execution of the training movement sequence, particularly, but not necessarily exclusively, information acquired by the sensor unit. Such adaptation of the machine learning model is achieved, for example, through backpropagation to update the model's parameters.

[0035] According to another advantageous embodiment, the computing unit uses a biomechanical model of the person when executing the generative pre-trained machine learning model.

[0036] The biomechanical model of the individual is preferably tailored to that specific person in order to accurately represent and utilize their physical characteristics. This model can serve as a reference for establishing restrictions and / or for precisely adjusting the training movement sequence. Furthermore, it can serve as a reference during the execution of the training movement sequence by the individual, particularly when a movement pattern of the training sequence is predefined on the model. Due to the high physiological similarity of the model to the individual, this allows for a comparison with the individual's actual movement.Accordingly, the model can be used to create a digital twin of the person as an athlete, in order to optimize and automatically determine a training movement sequence using the digital twin, and to provide feedback to the person on the actual execution of the determined training movement sequence compared to the determined one, especially in real time during the execution of the training movement sequence.

[0037] According to a further advantageous embodiment, data is recorded by the sensor unit during the execution of the training movement sequence by the person, wherein the computing unit uses the data of the execution of the training movement sequence to check whether those muscle fibers are currently fatigued which are also primarily used in the execution of the target movement sequence, and wherein the computing unit issues a warning if fatigue is detected.

[0038] According to a further advantageous embodiment, data is recorded by the sensor unit during the execution of the training movement sequence by the person, wherein the computing unit provides real-time feedback on the difference in the number of active motor neurons or muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence, based on the data of the execution of the training movement sequence by means of an output unit of the person.

[0039] This allows unnecessary training of uninvolved muscle groups to be stopped, and the individual's training status can be precisely optimized for the required target movement. A key advantage is that the deep insight into muscle mechanics provided by high-density electromyography can be displayed in real time during the execution of the training movement sequence via feedback. This allows the individual to be informed if muscles or specific fibers are being overtrained, or if the desired motor neurons are not being activated correctly.

[0040] According to a further advantageous embodiment, the computing unit provides the person with real-time feedback via an output unit during the execution of the training movement sequence determined by the computing unit, indicating whether the actual movement sequence matches the determined training movement sequence.

[0041] To verify the conformity of the actual movement sequence with the determined training movement sequence, the actual movement sequence is preferably recorded using a motion capture device such as an optical motion capture system. Alternatively, an inertial measurement unit attached to the person's body can also be used to record the person's actual movement sequence.

[0042] According to a further advantageous embodiment, data is recorded by the sensor unit during the execution of the training movement sequence by the person, wherein a recommended pause time is determined by the computing unit using the data of the execution of the training movement sequence at the end of the training movement sequence and is output by means of an output unit.

[0043] Another aspect of the invention relates to a system for the automated optimization of sports training, comprising a wearable, non-invasive high-density electromyography sensor unit designed to determine which motor neurons are active during a movement of the person, further comprising a processing unit designed to acquire the data from the sensor unit, and for evaluating the data acquired by the sensor unit during a sports target movement sequence of the person to be optimized, in order to obtain information about active motor neurons or about muscle fibers of the person activated by the motor neurons during the target movement sequence, and to store the information in a personal database, and based on the stored information to determine a training movement sequence for the person that is different from the target movement sequence.However, according to a pre-stored muscle-building model, there is a minimal difference in terms of active motor neurons or muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence.

[0044] Advantages and preferred further developments of the proposed system result from an analogous and substantive transfer of the above statements made in connection with the proposed procedure.

[0045] Further advantages, features and details will become apparent from the following description, in which - possibly with reference to the drawing - at least one embodiment is described in detail.

[0046] They show: Fig. 1: A method for the automated optimization of a sports training program according to an embodiment of the invention. Fig. 2: An exemplary target movement sequence and a corresponding exemplary training movement sequence in the procedure of Fig. 1. Fig. 3: A system for the automated optimization of a sports training program according to an embodiment of the invention.

[0047] The representations in the figures are schematic and not to scale.

[0048] Fig. Figure 1 shows a method for the automated optimization of sports training. A person, a top athlete, wears a wearable, non-invasive high-density electromyography sensor unit 1 as a leg covering S1 of their trousers. The athlete is, for example, a track and field sprinter and performs a sprint run with this sensor unit 1 attached to their body, representing the athletic target movement sequence S2 to be optimized. The start of such a sprint run, in the sense of the target movement sequence, is shown in the Fig. 2 is outlined in partial diagram (A). The aim is not simply for the athlete to perform the sprint in their training, but to specifically strengthen the muscle groups and muscle fibers used for sprinting with a training movement sequence. For this purpose, during the target movement sequence, sensor unit 1 records data on muscle activity with respect to the fibers (S3), and a processing unit 2 evaluates the data from sensor unit 1 for information about active motor neurons or muscle fibers activated by the motor neurons during the target movement sequence (S4), and the information is stored in a personal database (S5) so that it can be assigned to the individual again in a later training session and is available for comparison.Based on stored information, processing unit 2 automatically determines a training movement sequence S6 for the individual. This sequence differs from the target movement sequence but, according to a pre-stored muscle-building model, exhibits a minimal difference in terms of active motor neurons or muscle fibers activated by motor neurons between the training and target movement sequences. The training movement sequence is generated by processing unit 2 using a machine learning model and designed to be performed by a person in a training environment such as a gym. Corresponding rules are implemented in the machine learning model, such as a predefined set of dumbbells, equipment, mats, and other aids.If the person's training movement sequence is known and specified and traceable using an output unit 3, the person performs the training movement sequence, continuing with the worn sensor unit 1. Such a training movement sequence, which leads to the sprint according to partial image (A) of the . Fig. 2 belongs, for example, to the deadlift, shown in partial image (B) of the Fig. 2. This training movement sequence is monitored by sensor unit 1, which also checks which motor neurons are active. The person receives feedback on whether they are performing the training movement sequence as intended, in particular whether the activation of the motor neurons sufficiently corresponds to that of the target movement sequence. Furthermore, via output unit 3, the processing unit 2 provides the person with feedback on how they may need to modify the actual execution of the training movement sequence to adjust the activation of the motor neurons to that of the target movement sequence. This occurs in real time, so that during the execution of the training movement sequence, the person can optimally train the muscle fibers required for the target movement sequence.Using motion tracking, processing unit 2 can compare the actual movement sequence with the predefined training sequence and thus provide more accurate feedback. Further sensor fusions are possible, in particular heart rate measurement, blood pressure measurement, blood oxygen saturation measurement, lactate measurement, the use of measurements from an inertial measurement unit on the person's body, or other suitable sensors.

[0049] Fig.Figure 3 shows a system for the automated optimization of sports training, comprising a non-invasive high-density electromyography sensor unit 1, wearable by a person as a trouser leg, which is designed to determine which alpha motor neurons are active during a movement of the person, further comprising a processing unit 2, which is designed to acquire the data from the sensor unit 1, and to evaluate the data acquired by the sensor unit 1 during a sports target movement sequence of the person to be optimized, in order to obtain information about active motor neurons or about muscle fibers of the person activated by the motor neurons during the target movement sequence and to store the information in a personal database, and to determine a training movement sequence for the person based on the stored information which is different from the target movement sequence.However, according to a pre-stored muscle-building model, there is a minimal difference in terms of active motor neurons or muscle fibers activated by motor neurons between the training movement sequence and the target movement sequence. Feedback on the conformity of the person's actual movement sequence with the specified training movement sequence is displayed, for example, on screen 3 of the person.

[0050] Preferably, the person receives haptic and / or tactile feedback from the processing unit 2, particularly preferably by arranging haptic and / or tactile feedback actuators, such as vibration motors, on a fabric component of the sensor unit 1. Preferably, the person is given an indication, through appropriate activation of the vibration motors, of the direction in which they must shift the position of the relevant body part and / or by what angle they must rotate the relevant body part in order to optimally follow the prescribed training movement sequence. For this purpose, the processing unit can also employ a motion tracking system in sensor fusion with the sensor unit 1 to achieve the activation of the motor neurons as intended in the training movement sequence.

Claims

[1] Method for the automated optimization of sports training, wherein a portable non-invasive high-density electromyography sensor unit (1) is attached to a person (S1), the person with the sensor unit (1) performs a sports target movement sequence to be optimized (S2), data are recorded by the sensor unit (1) during the target movement sequence (S3), the data from the sensor unit (1) are evaluated by a computing unit (2) for information about active motor neurons or about muscle fibers of the person activated by the motor neurons during the target movement sequence (S4), and the information is stored in a personal database (S5);and wherein the computing unit (2) determines a training movement sequence (S6) for the person based on the stored information, which is different from the target movement sequence, but exhibits a minimal difference between the training movement sequence and the target movement sequence with respect to active motor neurons or muscle fibers activated by the motor neurons according to a pre-stored muscle-building model, wherein the automatic determination by the computing unit (2) is carried out in such a way that the training movement sequence is maximized to train precisely those muscle fibers which are explicitly used in the target movement sequence according to the database, or are implicitly considered to be activated by means of the identified predominantly active motor neurons.; [2] Method according to claim 1, wherein the training movement sequence is selected by the computing unit (2) as the one from a plurality of pre-stored possible training movement sequences which has a minimal difference with respect to the active motor neurons or the muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence. [3] Method according to claim 1, wherein the training movement sequence is assembled by the computing unit (2) by a combination of pre-stored training movement sequence elements such that the resulting training movement sequence has a minimal difference with respect to the active motor neurons or the muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence. [4] Method according to claim 1, wherein the training movement sequence is generated by the computing unit (2) by applying a generative pre-trained machine learning model. [5] Method according to claim 4, wherein the computing unit (2) uses a biomechanical model of the person when executing the generative pre-trained model of machine learning. [6] Method according to one of claims 4 to 5, wherein the machine learning model is adapted by the computing unit (2) during the execution of the training movement sequence in order to minimize the difference in terms of active motor neurons or muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence. [7] A method according to one of the preceding claims, wherein the sensor unit (1) records data during the execution of the training movement sequence by the person, wherein the computing unit (2) uses the data of the execution of the training movement sequence to check whether those muscle fibers are currently fatigued which are also primarily used in the execution of the target movement sequence, and wherein the computing unit (2) issues a warning when fatigue is detected. [8] Method according to one of the preceding claims, wherein data is recorded by the sensor unit (1) during the execution of the training movement sequence by the person, wherein the computing unit (2) provides real-time feedback to the person, based on the data of the execution of the training movement sequence, by means of an output unit (3) during the execution of the training movement sequence determined by the computing unit (2), regarding the difference in the number of active motor neurons or muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence. [9] Method according to one of the preceding claims, wherein the computing unit (2) provides the person with real-time feedback via an output unit (3) during the execution of the training movement sequence determined by the computing unit (2) regarding the conformity of an actual movement sequence with the determined training movement sequence. [10] Method according to one of the preceding claims, wherein data is recorded by the sensor unit (1) during the execution of the training movement sequence by the person, wherein a recommended pause time is determined by the computing unit (2) with the data of the execution of the training movement sequence at the end of the training movement sequence and is output by means of an output unit (3). [11] System for the automated optimization of sports training, comprising a wearable, non-invasive high-density electromyography sensor unit (1) designed to determine which motor neurons are active during a movement of the person, further comprising a computing unit (2) designed to acquire the data from the sensor unit (1), and for evaluating the data acquired by the sensor unit (1) during a sports target movement sequence of the person to be optimized, for information about active motor neurons or about muscle fibers of the person activated by the motor neurons during the target movement sequence, and for storing the information in a personal database, and for determining a training movement sequence for the person based on the stored information which is different from the target movement sequence,However, according to a pre-stored muscle-building model, there is a minimal difference in terms of active motor neurons or muscle fibers activated by the motor neurons between the training movement sequence and the target movement sequence.

Images