Sports injury risk detection method and system and sports injury risk part positioning method and system

Abstract

The invention discloses an exercise injury risk detection method and system and an exercise injury risk part positioning method and system. The method comprises the following steps: collecting multiple types of biological signals of a human body at a daily preset time point and preset time points before and after target exercise training to form an original signal data set; performing unified coding on the original signal data set to obtain a biological signal coding sequence; constructing a Gaussian noise adversarial generative network model, wherein the Gaussian noise adversarial generative network model comprises a generator and a discriminator; inputting the biological signal coding sequence into a Gaussian generative adversarial network model, and setting a total optimization target for adversarial training to obtain a trained generator and a discriminator; and taking the trained discriminator as an exercise injury risk detection model, obtaining a to-be-identified human body biological signal, inputting the to-be-identified human body biological signal into the exercise injury risk detection model, and obtaining an exercise injury risk detection result. The method and the model can objectively detect and position the sick part of the sports, and the detection accuracy is high.

Description

Technical Field

[0001] The present invention relates to the field of sports injury detection, and more specifically, to a sports injury risk detection method and system, and a sports injury risk location location method and system. Background Art

[0002] With the continuous improvement of sports competition level and the increasing performance requirements for athletes, how to effectively prevent injuries has become one of the key issues of concern in the sports industry. Many athletes have no scientific injury risk warnings, so their training methods are unscientific or they participate in high-intensity competitions with high injury risks, resulting in serious injuries, greatly shortening their careers, and even endangering their lives and health.

[0003] Traditionally, the detection of sports injuries mainly relies on the athletes' own reports and the experience and judgment of the medical team. This method has the disadvantages of strong subjectivity and obvious lag. In addition, due to the large differences in physical condition, technical level and training load of different athletes in different fields, it is difficult to meet personalized needs by adopting a unified standard, which greatly reduces the accuracy and timeliness of injury prediction. There are also some methods that use statistical methods or machine learning algorithms to detect sports injuries. Although they can reflect the changing trends of physiological parameters to a certain extent, it is difficult to capture subtle changes in physiological parameters and the potential injury risks behind them; at the same time, such models cannot predict the current players' injury risks in real time online, which makes it impossible for athletes and coaching staff to deal with injuries and accidents that occur in real time on the field.

[0004] The prior art discloses a method, device, and medium for alternating training and medical image classification of a generative adversarial network, which measures the convergence degree of the generator and the discriminator, and adaptively determines whether to train the generator or the discriminator next time based on the convergence degree of the current generator and the discriminator; finally, the medical images are classified for disease through an adaptive generative adversarial network. This method mainly relies on medical images for classification, and does not combine multiple biological signals, so it is difficult to capture subtle changes and the potential injury risks behind them. At the same time, this method is complex to train and cannot predict the current players' injury risks online in real time. Summary of the invention

[0005] The present invention aims at the defects of the prior art that the injury detection accuracy is low and the injury site cannot be reflected; and provides a sports injury risk detection method and system and a sports injury risk site positioning method and system. The method can objectively detect and locate the sports injury site with high detection accuracy.

[0006] The primary purpose of the present invention is to solve the above technical problems. The technical solution of the present invention is as follows:

[0007] A sports injury risk detection method, comprising:

[0008] S101: Collecting multiple types of biological signals of the human body at daily preset time points and preset time points before and after target exercise training to form an original signal data set;

[0009] S102: uniformly encoding the original signal data set to obtain a biological signal coding sequence;

[0010] S103: Construct a Gaussian noise adversarial generative network model, including a generator and a discriminator;

[0011] S104: inputting the biological signal coding sequence into a Gaussian adversarial generative network model, setting an overall optimization target for adversarial training, and obtaining a trained generator and discriminator;

[0012] S105: Using the trained discriminator as a sports injury risk detection model, obtaining a human biological signal to be identified, inputting the signal into the sports injury risk detection model, and obtaining a sports injury risk detection result.

[0013] Furthermore, in step S102, the original signal data set is uniformly encoded, including:

[0014] S10201: resampling multiple types of biological signals in the original signal data set to obtain a resampled original signal data set;

[0015] S10202: segmenting the biological signals in the resampled original signal data set according to the same preset time window to obtain a corresponding original signal matrix sequence;

[0016] S10203: Calculate the position coordinate code and electrode channel code matrix of each original signal matrix in each original signal matrix sequence;

[0017] S10204: unifying the dimensions of the original signal matrix sequence to obtain a corresponding original signal matrix sequence after unification of the dimensions;

[0018] S10205: adding corresponding position coordinate codes and corresponding electrode channel code matrices to each original signal matrix in the original signal matrix sequence after each dimension is unified, to obtain the original signal matrix after adding corresponding position coordinate codes and electrode channel codes;

[0019] S10206: splice the original signal matrix after the added position coordinate coding and electrode channel coding corresponding to the same time to obtain a biological signal coding sequence.

[0020] Furthermore, in step S10203, the formula for encoding the position coordinates is as follows:

[0021]

[0022] t represents the unit time serial number, i represents the number of electrode channels, and dim represents the maximum number of electrode channels in the original signal matrix sequence after time division.

[0023] Furthermore, in step S10203, the calculation of the electrode channel coding matrix includes:

[0024] S1020301: taking an average value of each original signal matrix according to the electrode channel to obtain a corresponding electrode channel average matrix;

[0025] S1020302: Unify the average matrix dimensions of each electrode channel to obtain a corresponding electrode channel encoding matrix.

[0026] Further, the generator includes a first perceptual machine layer, a first noise perturbation layer, a second perceptual machine layer, a second noise perturbation layer, a third perceptual machine layer, a third noise perturbation layer, a fourth perceptual machine layer, a fourth noise perturbation layer, a fifth perceptual machine layer, and a fifth noise perturbation layer arranged in sequence;

[0027] The discriminator includes a first convolutional layer, a first BN layer, a second convolutional layer, a second BN layer, a third convolutional layer, a third BN layer, a flattening layer, and a fully connected layer, which are arranged in sequence.

[0028] Furthermore, the overall optimization goal is:

[0029]

[0030] G represents the generator, D represents the discriminator, V(D,G) represents the loss function, E represents the expectation, P x (x) represents the input distribution, N P represents noise, N(x) represents noise distribution, and x represents input;

[0031] The loss function of the generator is:

[0032]

[0033] E stands for expectation, N P represents noise, N(x) represents noise distribution, x represents input, P x (x) represents the input distribution, G represents the generator, and D represents the discriminator;

[0034] The loss function of the discriminator is:

[0035]

[0036] E stands for expectation, N Prepresents noise, N(x) represents noise distribution, x represents input, P x (x) represents the input distribution, G represents the generator, and D represents the discriminator.

[0037] A sports injury risk detection system, comprising:

[0038] Original signal acquisition module: collects multiple types of biological signals of the human body at preset time points in daily life and at preset time points before and after target exercise training to form an original signal data set;

[0039] Original signal encoding module: uniformly encodes the original signal data set to obtain a biological signal encoding sequence;

[0040] Adversarial Generative Network Model Building Module: Building a Gaussian noise adversarial generative network model, including a generator and a discriminator;

[0041] Adversarial training module: inputting the biological signal coding sequence into the Gaussian adversarial generative network model, setting the overall optimization target for adversarial training, and obtaining the trained generator and discriminator;

[0042] Injury risk detection module: Use the trained discriminator as a sports injury risk detection model, obtain the human biological signal to be identified, input the sports injury risk detection model, and obtain the sports injury risk detection result.

[0043] A method for locating risk areas of sports injuries, comprising:

[0044] S201: Collecting multiple types of biological signals of multiple target parts of the human body at preset time points after target movement training to form a disease source localization data set;

[0045] S202: modifying the sports injury risk detection model to obtain a sports injury risk feature extraction model;

[0046] S203: inputting the pathogen location data set into the sports injury risk feature extraction model to obtain corresponding pathogen risk features;

[0047] S204: Clustering the pathogen risk characteristics to obtain multiple final cluster centers;

[0048] S205: obtaining a human biological signal to be located, inputting it into a trained sports injury risk feature extraction model, and obtaining the injury risk feature to be located;

[0049] S206: Calculate the Mahalanobis distance between the injury risk feature with location and each final cluster center respectively and sort them, and obtain the cluster center corresponding to the minimum Mahalanobis distance as the minimum cluster center;

[0050] S207: The target part corresponding to the pathogenic risk feature with the largest proportion in the cluster corresponding to the minimum cluster center is used as the sports injury risk part.

[0051] Furthermore, the sports injury risk feature extraction model removes the fully connected layer in the sports injury risk detection model.

[0052] A sports injury risk location system, comprising:

[0053] Pathogen localization data collection module: collects multi-type biological signals of multiple target parts of the human body at preset time points after target movement training to form a disease source localization data set;

[0054] Feature extraction model construction module: modify the sports injury risk detection model to obtain a sports injury risk feature extraction model;

[0055] Pathogen risk feature extraction module: input the pathogen location data set into the sports injury risk feature extraction model to obtain the corresponding pathogen risk feature;

[0056] Clustering module: clustering the pathogen risk characteristics to obtain multiple final clustering centers;

[0057] Injury risk feature extraction module: obtains the biological signal of the human body to be located, inputs the trained sports injury risk feature extraction model, and obtains the injury risk feature to be located;

[0058] Mahalanobis distance calculation module: calculates the Mahalanobis distance between the injury risk characteristics with location and each final cluster center and sorts them, and obtains the cluster center corresponding to the minimum Mahalanobis distance as the minimum cluster center;

[0059] Target site positioning module: The target site corresponding to the pathogen risk feature with the largest proportion in the cluster corresponding to the minimum cluster center is taken as the sports injury risk site.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] The present invention collects multiple types of biological signals of the human body at daily preset time points and preset time points before and after target sports training to form an original signal data set; uniformly encodes the original signal data set to obtain a biological signal coding sequence; constructs a Gaussian noise adversarial generative network model, including a generator and a discriminator; inputs the biological signal coding sequence into the Gaussian adversarial generative network model, sets an overall optimization goal for adversarial training, and obtains a trained generator and discriminator; uses the trained discriminator as a sports injury risk detection model to obtain a human biological signal to be identified, inputs it into the sports injury risk detection model, and obtains a sports injury risk detection result; the sports injury risk can be objectively detected with high detection accuracy.

[0062] By collecting multi-type biological signals of multiple target parts of the human body at preset time points after target sports training, a pathogen location data set is formed; the sports injury risk detection model is modified to obtain a sports injury risk feature extraction model; the pathogen location data set is input into the sports injury risk feature extraction model to obtain the corresponding pathogen risk features; the pathogen risk features are clustered to obtain multiple final cluster centers; the biological signals of the human body to be located are obtained, and input into the trained sports injury risk feature extraction model to obtain the injury risk features to be located; the Mahalanobis distances between the located injury risk features and each final cluster center are calculated and sorted, and the cluster center corresponding to the minimum Mahalanobis distance is obtained as the minimum cluster center; the target part corresponding to the pathogen risk feature with the largest proportion in the cluster corresponding to the minimum cluster center is used as the sports injury risk part; thereby accurately locating the sports injury risk position. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a flow chart of a sports injury risk detection method provided in Example 1.

[0064] Figure 2 This is a flow chart of the method for collecting the original signal data set provided in Example 1.

[0065] Figure 3 The flowchart of uniformly encoding the original signal data set provided in Example 1.

[0066] Figure 4 A schematic diagram of the principle of uniformly encoding the original signal data set provided in Example 1.

[0067] Figure 5 This is a flow chart of the calculation of the electrode channel encoding matrix provided in Example 1.

[0068] Figure 6 This is a structural diagram of the generator provided in Example 1.

[0069] Figure 7 This is a structural diagram of the discriminator provided in Example 1.

[0070] Figure 8 This is a training flowchart of the Gaussian adversarial generative network model provided in Example 1.

[0071] Fig. 9 This is a structural diagram of a sports injury risk detection system provided in Example 1.

[0072] Fig.10 This is a flow chart of a method for locating sports injury risk areas provided in Example 1.

[0073] Fig.11 This is a structural diagram of a sports injury risk location system provided in Example 1.

[0074] Fig.12 This is a flow chart of a real-time sports injury risk detection method provided in Example 2. DETAILED DESCRIPTION

[0075] The drawings are for illustrative purposes only and should not be construed as limiting the present patent;

[0076] In order to better illustrate the present embodiment, some parts in the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product;

[0077] It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0078] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0079] Example 1

[0080] like Figure 1 As shown, a method for detecting sports injury risk includes:

[0081] S101: Collecting multiple types of biological signals of the human body at daily preset time points and preset time points before and after target exercise training to form an original signal data set;

[0082] S102: uniformly encoding the original signal data set to obtain a biological signal coding sequence;

[0083] S103: Construct a Gaussian noise adversarial generative network model, including a generator and a discriminator;

[0084] S104: inputting the biological signal coding sequence into a Gaussian adversarial generative network model, setting an overall optimization target for adversarial training, and obtaining a trained generator and discriminator;

[0085] S105: Using the trained discriminator as a sports injury risk detection model, obtaining a human biological signal to be identified, inputting the signal into the sports injury risk detection model, and obtaining a sports injury risk detection result.

[0086] In a specific embodiment, if the sports injury risk detection result is greater than a preset threshold, preferably 0.6, it indicates that the athlete's physical condition may have reached its limit and continued training or competition is likely to lead to injury.

[0087] In a specific embodiment, two biological signals are used as an example: electromyogram and electrocardiogram. First, the user's different periods are collected by electromyography and electrocardiography, such as the three periods of morning, noon and evening, three minutes before training, three minutes after training, or the period after a certain special training (such as sprint training of short-distance running, etc.), and these biological signals form the original signal data set.

[0088] In a specific embodiment, Figure 2 As shown, for the field of basketball, six periods are selected: 8:00 and 11:00 in the morning, 13:00 and 15:00 in the afternoon, and 18:00 and 20:00 in the evening, as well as special training programs for the field of basketball: eight basic skill trainings: powerful dribbling on the spot, dribbling over obstacles, spot shooting, pull-up jump shot, rebound pass, running layup, sliding step and crossover step; five endurance trainings: 5000-meter long-distance running, sprint running, box jumps, squat jumps and quick reaction; eight tactical and psychological quality trainings: meditation, breathing control, man-to-man defense, full-court pressing, one-on-one offense and defense exercises, fast break, pick-and-roll, and triangle offense.

[0089] Three-minute electrocardiogram, electromyogram, and electroencephalogram signals of the athletes were collected 2 minutes before and 2 minutes after the aforementioned training, as well as in the six periods mentioned above, to construct a database of original signals.

[0090] Furthermore, if Figure 3 , Figure 4 As shown, in step S102, the original signal data set is uniformly encoded, including:

[0091] S10201: resampling multiple types of biological signals in the original signal data set to obtain a resampled original signal data set;

[0092] S10202: segmenting the biological signals in the resampled original signal data set according to the same preset time window to obtain a corresponding original signal matrix sequence;

[0093] S10203: Calculate the position coordinate code and electrode channel code matrix of each original signal matrix in each original signal matrix sequence;

[0094] S10204: unifying the dimensions of the original signal matrix sequence to obtain a corresponding original signal matrix sequence after unification of the dimensions;

[0095] S10205: adding corresponding position coordinate codes and corresponding electrode channel code matrices to each original signal matrix in the original signal matrix sequence after each dimension is unified, to obtain the original signal matrix after adding corresponding position coordinate codes and electrode channel codes;

[0096] S10206: splice the original signal matrix after the added position coordinate coding and electrode channel coding corresponding to the same time to obtain a biological signal coding sequence.

[0097] It should be noted that since the sampling rates of the various signals are different, a resampling operation is required to unify the sampling rates. Since the number of channels of the various signals is different, the dimensions need to be unified to facilitate subsequent processing.

[0098] In a specific embodiment, dimensionality unification expands a signal with a smaller dimension to the dimension of a signal with a larger dimension.

[0099] Furthermore, in step S10203, the formula for encoding the position coordinates is as follows:

[0100]

[0101] t represents the unit time serial number, i represents the number of electrode channels, and dim represents the maximum number of electrode channels in the original signal matrix sequence after time division.

[0102] Furthermore, if Figure 5 As shown, in step S10203, the calculation of the electrode channel coding matrix includes:

[0103] S1020301: taking an average value of each original signal matrix according to the electrode channel to obtain a corresponding electrode channel average matrix;

[0104] S1020302: Unify the average matrix dimensions of each electrode channel to obtain a corresponding electrode channel encoding matrix.

[0105] Furthermore, if Figure 6 As shown, the generator includes a first perceptual machine layer, a first noise perturbation layer, a second perceptual machine layer, a second noise perturbation layer, a third perceptual machine layer, a third noise perturbation layer, a fourth perceptual machine layer, a fourth noise perturbation layer, a fifth perceptual machine layer, and a fifth noise perturbation layer arranged in sequence;

[0106] like Figure 7As shown, the discriminator includes a first convolutional layer, a first BN layer, a second convolutional layer, a second BN layer, a third convolutional layer, a third BN layer, a flattening layer, and a fully connected layer arranged in sequence.

[0107] It should be noted that different Gaussian noise inputs are added to each layer, and the output is perturbed in multiple layers to simulate injuries. In this way, the output of the generator will be as consistent as possible with the fluctuation distribution of the original output, but there will be obvious perturbation noise to achieve accurate simulation of the injury signal.

[0108] In a specific embodiment, the first noise disturbance layer, the second noise disturbance layer, the third noise disturbance layer, and the fourth noise disturbance layer add noise four times respectively, and the added intensities are 20%, 30%, 40%, and 50% respectively.

[0109] Furthermore, the overall optimization goal is:

[0110]

[0111] G represents the generator, D represents the discriminator, V(D,G) represents the loss function, E represents the expectation, P x (x) represents the input distribution, N P represents noise, N(x) represents noise distribution, and x represents input;

[0112] The loss function of the generator is:

[0113]

[0114] E stands for expectation, N P represents noise, N(x) represents noise distribution, x represents input, P x (x) represents the input distribution, G represents the generator, and D represents the discriminator;

[0115] The loss function of the discriminator is:

[0116]

[0117] E stands for expectation, N P represents noise, N(x) represents noise distribution, x represents input, P x (x) represents the input distribution, G represents the generator, and D represents the discriminator.

[0118] The training process of the Gaussian adversarial network model is as follows Figure 8 shown.

[0119] like Fig. 9 As shown, a sports injury risk detection system includes:

[0120] Original signal acquisition module: collects multiple types of biological signals of the human body at preset time points in daily life and at preset time points before and after target exercise training to form an original signal data set;

[0121] Original signal encoding module: uniformly encodes the original signal data set to obtain a biological signal encoding sequence;

[0122] Adversarial Generative Network Model Building Module: Building a Gaussian noise adversarial generative network model, including a generator and a discriminator;

[0123] Adversarial training module: inputting the biological signal coding sequence into the Gaussian adversarial generative network model, setting the overall optimization target for adversarial training, and obtaining the trained generator and discriminator;

[0124] Injury risk detection module: Use the trained discriminator as a sports injury risk detection model, obtain the human biological signal to be identified, input the sports injury risk detection model, and obtain the sports injury risk detection result.

[0125] like Fig.10 As shown, a method for locating risk areas of sports injuries includes:

[0126] S201: Collecting multiple types of biological signals of multiple target parts of the human body at preset time points after target movement training to form a disease source localization data set;

[0127] S202: modifying the sports injury risk detection model to obtain a sports injury risk feature extraction model;

[0128] S203: inputting the pathogen location data set into the sports injury risk feature extraction model to obtain corresponding pathogen risk features;

[0129] S204: Clustering the pathogen risk characteristics to obtain multiple final cluster centers;

[0130] S205: obtaining a human biological signal to be located, inputting it into a trained sports injury risk feature extraction model, and obtaining the injury risk feature to be located;

[0131] S206: Calculate the Mahalanobis distance between the injury risk feature with location and each final cluster center respectively and sort them, and obtain the cluster center corresponding to the minimum Mahalanobis distance as the minimum cluster center;

[0132] S207: The target part corresponding to the pathogenic risk feature with the largest proportion in the cluster corresponding to the minimum cluster center is used as the sports injury risk part.

[0133] In a specific embodiment, the electromyographic signals and electroencephalographic signals are specifically recorded after three minutes of special training of body parts such as leg training, arm muscle training, and abdominal training to construct a disease source localization data set.

[0134] In a specific embodiment, in the field of basketball, nine parts of the body that are very prone to injury, namely ankles, knees, fingers, elbows, shoulders, head, thigh muscles, biceps, and abdominal muscles, are selected and subjected to a certain low-intensity stress training, i.e., the parts are forced to be used for one minute. The electrocardiogram, electromyogram, and electroencephalogram of the athletes are recorded 2 minutes before and 2 minutes after the stress training, respectively, to construct a disease source location database.

[0135] In a specific embodiment, clustering uses K-Means clustering, including:

[0136] S20401: Randomly select N pathogen risk features as N initial cluster centers;

[0137] S20402: Assign each pathogen risk feature to the cluster closest to the cluster center;

[0138] S20403: Update the cluster center of each of the N clusters;

[0139] S20404: Repeat steps S20402 to S20403 until the cluster center no longer changes or a preset number of iterations is reached;

[0140] S20405: Take the cluster center as the final cluster center.

[0141] In a specific embodiment, since low-intensity stress training is performed on 9 body parts in total, the number of clusters is specified to be 9, and 9 cluster centers can be obtained.

[0142] The formula for updating the cluster center is:

[0143]

[0144] C(k) represents the k-th cluster set, |C(k)| represents the number of pathogenic features in the k-th cluster, i represents the sequence number, X outputware,i represents the risk characteristic of the i-th pathogen.

[0145] In a specific embodiment, the calculation formula of the Mahalanobis distance is as follows:

[0146]

[0147] Δ is the covariance matrix of all final cluster centers and injury risk characteristics, V outputbs represents the injury risk characteristics, μ center represents the final cluster center to be calculated, Δ-1 represents the inverse matrix of Δ.

[0148] It should be noted that the Mahalanobis distance takes into account the covariance matrix of the vector and can better reflect the correlation between data points in high-dimensional space.

[0149] Furthermore, the sports injury risk feature extraction model removes the fully connected layer in the sports injury risk detection model.

[0150] like Fig.11 As shown, a sports injury risk location system includes:

[0151] Pathogen localization data collection module: collects multi-type biological signals of multiple target parts of the human body at preset time points after target movement training to form a disease source localization data set;

[0152] Feature extraction model construction module: modify the sports injury risk detection model to obtain a sports injury risk feature extraction model;

[0153] Pathogen risk feature extraction module: input the pathogen location data set into the sports injury risk feature extraction model to obtain the corresponding pathogen risk feature;

[0154] Clustering module: clustering the pathogen risk characteristics to obtain multiple final clustering centers;

[0155] Injury risk feature extraction module: obtains the biological signal of the human body to be located, inputs the trained sports injury risk feature extraction model, and obtains the injury risk feature to be located;

[0156] Mahalanobis distance calculation module: calculates the Mahalanobis distance between the injury risk characteristics with location and each final cluster center and sorts them, and obtains the cluster center corresponding to the minimum Mahalanobis distance as the minimum cluster center;

[0157] Target site positioning module: The target site corresponding to the pathogen risk feature with the largest proportion in the cluster corresponding to the smallest cluster center is taken as the sports injury risk site.

[0158] Example 2

[0159] Based on the sports injury risk detection method described in Example 1, that is, this embodiment adopts the same sports injury risk detection method as Example 1.

[0160] like Fig.12As shown in the figure, in order to improve the practical effect of the model, the test-time adaptation (TTA) technology is combined to add the function of real-time detection of athletes' injury risks. The model is trained in advance using the data collected during the daily training phase to obtain a customized pre-trained model dedicated to the athlete. In the competition and other stages where real-time detection is required, we put simple sensors on the players, which can obtain the athletes' various biological signals in real time, and make real-time sports injury risk detection and injury risk location.

[0161] Since athletes wear simple and lightweight sensors during the competition, and the competition status and training status are different, the distribution of biological signals also changes. If the discriminator described in Example 1 is used directly, deviations may occur. Therefore, if real-time testing is required, the first BN layer, the second BN layer, and the third BN layer use a sliding average method to modify the statistics.

[0162] The principle of modifying statistics in a sliding average manner is as follows:

[0163] The first BN layer, the second BN layer, and the third BN layer in Example 1 are normalized using the global mean and variance, and the formula is as follows:

[0164]

[0165] represents the input of the i-th BN layer, It represents the first processing result of the input of the i-th BN layer. represents the second processing result of the input of the i-th BN layer, μ and σ 2 are the accumulated global mean and variance respectively, ∈ is a very small value to prevent the denominator from being zero, γ and β are learnable parameters, and i represents the sequence number.

[0166] When only one data point is passed in at a time, the batch size is too small to accurately estimate the mean and variance.

[0167] Therefore, the improved formulas for the mean and variance are as follows:

[0168]

[0169] μ ′ represents the improved mean, σ ′2 Represents the improved variance, α represents the momentum parameter, x represents the i-th input, and i represents the sequence number.

[0170] It should be noted that the moving average statistic μ ′ and σ ′2Normalization and scaling are performed. After each real-time signal is input and the test results are obtained, the model parameters will return to the original pre-trained state, which means that the tests between each signal segment are independent and do not affect each other. Each slight modification of the model parameters according to the incoming biological signal fragment can help eliminate the differences in signal distribution under different sensors or different occasions, and ensure that the original pre-trained model after a large amount of data training is still used to play a discriminative role, ensuring the training accuracy and eliminating the distribution error.

[0171] The same or similar reference numerals correspond to the same or similar components;

[0172] The terms used in the drawings to describe positional relationships are only used for illustrative purposes and should not be construed as limiting this patent;

[0173] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A sports injury risk detection method, characterized in that: include: S101: Collecting multiple types of biological signals of the human body at daily preset time points and preset time points before and after target exercise training to form an original signal data set; S102: uniformly encoding the original signal data set to obtain a biological signal coding sequence; S103: Construct a Gaussian noise adversarial generative network model, including a generator and a discriminator; S104: inputting the biological signal coding sequence into a Gaussian adversarial generative network model, setting an overall optimization target for adversarial training, and obtaining a trained generator and discriminator; S105: Using the trained discriminator as a sports injury risk detection model, obtaining a human biological signal to be identified, inputting the signal into the sports injury risk detection model, and obtaining a sports injury risk detection result.

2. A sports injury risk detection method according to claim 1, characterized in that: In step S102, the original signal data set is uniformly encoded, including: S10201: resampling multiple types of biological signals in the original signal data set to obtain a resampled original signal data set; S10202: segmenting the biological signals in the resampled original signal data set according to the same preset time window to obtain a corresponding original signal matrix sequence; S10203: Calculate the position coordinate code and electrode channel code matrix of each original signal matrix in each original signal matrix sequence; S10204: unifying the dimensions of the original signal matrix sequence to obtain a corresponding original signal matrix sequence after unification of the dimensions; S10205: adding corresponding position coordinate codes and corresponding electrode channel code matrices to each original signal matrix in the original signal matrix sequence after each dimension is unified, to obtain the original signal matrix after adding corresponding position coordinate codes and electrode channel codes; S10206: splice the original signal matrix after the added position coordinate coding and electrode channel coding corresponding to the same time to obtain a biological signal coding sequence.

3. A sports injury risk detection method according to claim 2, characterized in that: In step S10203, the formula for encoding the position coordinates is as follows: t represents the unit time serial number, i represents the number of electrode channels, and dim represents the maximum number of electrode channels in the original signal matrix sequence after time division.

4. A sports injury risk detection method according to claim 2, characterized in that: In step S10203, the calculation of the electrode channel coding matrix includes: S1020301: taking an average value of each original signal matrix according to the electrode channel to obtain a corresponding electrode channel average matrix; S1020302: Unify the average matrix dimensions of each electrode channel to obtain a corresponding electrode channel encoding matrix.

5. A sports injury risk detection method according to claim 1, characterized in that: The generator includes a first perceptual machine layer, a first noise perturbation layer, a second perceptual machine layer, a second noise perturbation layer, a third perceptual machine layer, a third noise perturbation layer, a fourth perceptual machine layer, a fourth noise perturbation layer, a fifth perceptual machine layer, and a fifth noise perturbation layer arranged in sequence; The discriminator includes a first convolutional layer, a first BN layer, a second convolutional layer, a second BN layer, a third convolutional layer, a third BN layer, a flattening layer, and a fully connected layer, which are arranged in sequence.

6. A sports injury risk detection method according to claim 1, characterized in that: The overall optimization goal is: G represents the generator, D represents the discriminator, V(D,G) represents the loss function, E represents the expectation, P x (x) represents the input distribution, N P represents noise, N(x) represents noise distribution, and x represents input; The loss function of the generator is: E stands for expectation, N P represents noise, N(x) represents noise distribution, x represents input, P x (x) represents the input distribution, G represents the generator, and D represents the discriminator; The loss function of the discriminator is: E stands for expectation, N P represents noise, N(x) represents noise distribution, x represents input, P x (x) represents the input distribution, G represents the generator, and D represents the discriminator.

7. A sports injury risk detection system, applied to the detection method according to any one of claims 1 to 6, characterized in that: include: Original signal acquisition module: collects multiple types of biological signals of the human body at preset time points in daily life and at preset time points before and after target exercise training to form an original signal data set; Original signal encoding module: uniformly encodes the original signal data set to obtain a biological signal encoding sequence; Adversarial Generative Network Model Building Module: Building a Gaussian noise adversarial generative network model, including a generator and a discriminator; Adversarial training module: inputting the biological signal coding sequence into the Gaussian adversarial generative network model, setting the overall optimization target for adversarial training, and obtaining the trained generator and discriminator; Injury risk detection module: Use the trained discriminator as a sports injury risk detection model, obtain the human biological signal to be identified, input the sports injury risk detection model, and obtain the sports injury risk detection result.

8. A method for locating risk areas of sports injuries, characterized in that: include: S201: collecting multiple types of biological signals of multiple target parts of the human body at preset time points after target movement training to form a disease source localization data set; S202: Modify the sports injury risk detection model as described in any one of claims 1 to 7 to obtain a sports injury risk feature extraction model; S203: Inputting the pathogen location data set into the sports injury risk feature extraction model to obtain corresponding pathogen risk features; S204: Clustering the pathogen risk characteristics to obtain multiple final clustering centers; S205: obtaining a human biological signal to be located, inputting it into a trained sports injury risk feature extraction model, and obtaining the injury risk feature to be located; S206: respectively calculating the Mahalanobis distance between the injury risk feature with location and each final cluster center and sorting them, obtaining the cluster center corresponding to the minimum Mahalanobis distance as the minimum cluster center; S207: The target part corresponding to the pathogenic risk feature with the largest proportion in the cluster corresponding to the minimum cluster center is used as the sports injury risk part.

9. A method for locating sports injury risk areas according to claim 8, characterized in that: The sports injury risk feature extraction model removes the fully connected layer in the sports injury risk detection model.

10. A sports injury risk location positioning system, applied to the positioning method according to claim 8 or 9, characterized in that: include: Pathogen localization data collection module: collects multi-type biological signals of multiple target parts of the human body at preset time points after target movement training to form a disease source localization data set; Feature extraction model construction module: modifying the sports injury risk detection model as described in any one of claims 1 to 7 to obtain a sports injury risk feature extraction model; Pathogen risk feature extraction module: input the pathogen location data set into the sports injury risk feature extraction model to obtain the corresponding pathogen risk feature; Clustering module: clustering the pathogen risk characteristics to obtain multiple final clustering centers; Injury risk feature extraction module: obtains the biological signal of the human body to be located, inputs the trained sports injury risk feature extraction model, and obtains the injury risk feature to be located; Mahalanobis distance calculation module: calculates the Mahalanobis distance between the injury risk characteristics with location and each final cluster center and sorts them, and obtains the cluster center corresponding to the minimum Mahalanobis distance as the minimum cluster center; Target site positioning module: The target site corresponding to the pathogen risk feature with the largest proportion in the cluster corresponding to the smallest cluster center is taken as the sports injury risk site.

Images