A real-time analysis system and method of muscle activity based on myoelectric signals

Abstract

The application discloses a real-time analysis system and method for muscle activity based on electromyographic signals, and relates to the technical field of electromyographic signals, and comprises the following steps: calling signal acquisition records of user project training, extracting corresponding sampling frequencies, and obtaining target sampling frequencies; collecting electromyographic data of muscle contraction completed by the user; decomposing the training project into a plurality of training actions, obtaining a reference vector set of each limb, and obtaining an effective time period of each training action; dividing the effective time period into a plurality of time windows, establishing a time domain function corresponding to the effective time period, and obtaining a target time length of each training action; acquiring a training action at the current training time, evaluating abnormal conditions of the training action, and giving early warning and prompt to relevant instructors. The application analyzes historical signal acquisition records in combination with multi-dimensional characteristics, monitors abnormal behaviors in the project training process in real time, gives early warning and prompt to instructors in a timely manner, and is helpful to reduce the risk of muscle injury of the user.

Description

Technical Field

[0001] This invention relates to the field of electromyography (EMG) signal technology, specifically to a real-time muscle activity analysis system and method based on EMG signals. Background Technology

[0002] Electromyography (EMG) signals include surface electromyography (sEMG) and inserted electromyography (iEMG). Surface electromyography (sEMG) is the sum of muscle electrical activity recorded from the skin surface. Essentially, it is the superposition of motor unit action potentials (MUAPs) generated by multiple recruited motor units on the body surface. As a non-invasive, real-time bioelectrical signal that reflects muscle activity, it has important application value in sports science, rehabilitation medicine, human-computer interaction and other fields.

[0003] In the field of sports science, dynamic monitoring of muscle fatigue and force patterns can prevent sports injuries and optimize technical movements. However, because existing research does not combine multi-dimensional features to analyze training movements, it is difficult to reflect the complexity of muscle synergistic activation, which leads to insufficient accuracy in movement recognition, delayed prompts for abnormal movements, and difficulty in meeting the needs of real-time detection and feedback, thus increasing the risk of muscle injury for users. Summary of the Invention

[0004] To address the aforementioned issues, this invention discloses a real-time muscle activity analysis system and method based on electromyographic signals, which meets the requirements for real-time detection and feedback, and helps reduce the risk of muscle injury for users.

[0005] A method for real-time analysis of muscle activity based on electromyographic signals includes the following steps:

[0006] Step 1: Retrieve the historical signal acquisition records of the user's project training, extract the sampling frequency corresponding to the signal acquisition records, and obtain the target sampling frequency based on the server's resource usage during the signal acquisition process; guide the user to complete the voluntary muscle contraction activity, and continuously acquire electromyographic data of the entire contraction process according to the target sampling frequency;

[0007] Step 2: Decompose the training program into several training movements and retrieve the baseline movement videos for each training movement; based on the positional relationship between adjacent limbs in the baseline movement videos, obtain the baseline vector set for each limb; based on the baseline vector set and the monitoring video corresponding to the electromyography data at the corresponding time, obtain the effective time period for each training movement.

[0008] Step 3: Obtain the minimum and maximum window durations of the time window, extract several window durations within the minimum and maximum window durations, divide the effective time period into several time windows, establish the time domain function corresponding to the effective time period, obtain the duration feature value of each window duration based on the time domain function, and obtain the target duration of each training action based on the duration feature value.

[0009] Step 4: Obtain the training movements during the current training session, obtain each time window based on the target duration of the training movements, evaluate any abnormalities in the training movements based on the frequency domain information within the time windows, and promptly issue warnings to relevant instructors.

[0010] Preferably, the continuous acquisition of electromyographic data throughout the entire contraction process includes:

[0011] Step 1: Retrieve the historical signal acquisition records of the user project training and extract the sampling frequency from each signal acquisition record; obtain the minimum sampling frequency preset by the system; and, based on the server's resource usage value during the signal acquisition process, use the sampling frequency at which the resource usage value equals the preset available resource usage value as the maximum sampling frequency.

[0012] The target sampling frequency is randomly selected from the lowest and highest sampling frequencies; the user is guided to complete the voluntary muscle contraction activity, and electromyographic data of the entire contraction process are continuously collected according to the target sampling frequency.

[0013] The target sampling frequency is determined here based on the minimum and maximum sampling frequencies, taking into account both system rules and resource consumption. Using the target sampling frequency to collect electromyography data from users can effectively improve system reliability.

[0014] Preferably, step 2 yields the effective time period for each training exercise, including:

[0015] Step 21: Decompose the training program into several training movements and retrieve the baseline movement videos for each training movement; obtain a time t from the baseline movement video corresponding to a certain training movement B, establish a 3D model, and obtain the position of each limb at time t. Take the adjacent limbs of a certain limb Z0 as Z1, and take the positions of the head and tail of limb Z0 at time t as P respectively. s 0 and P w 0 The positions of the head and tail of limb Z1 at time t are respectively designated as P. s 1 and P w 1 ;

[0016] Step 22: Set the number of limb segments to N, and obtain the positions P according to uniform length intervals. s0 and P w 0 There are N points within a given area, where N > 2, and the points include P. s 0 and P w 0 And sort the positions in order from beginning to end. Similarly, obtain the positions P according to uniform length intervals. s 1 and P w 1 There are N location points within the range, including point P. s 1 and P w 1 The positions are also sorted in order from beginning to end; two positions with the same index are obtained in limb Z0 and limb Z1, and the vector pointing from the position point in limb Z0 to the position point in limb Z1 is used as a reference vector of limb Z0, thus obtaining the reference vector set of limb Z0.

[0017] Step 23: Obtain the time corresponding to the electromyography data and retrieve the corresponding monitoring video; obtain the positions of limbs Z0 and Z1 at a certain time t0 in the monitoring video, obtain the actual vector set corresponding to limb Z0, the number of elements in the actual vector set is N, and obtain the deviation value: Where e is the natural constant, and V n W is the nth vector in the reference vector set. n The nth vector in the actual vector set; if the deviation value is less than the preset deviation threshold, the user's training action at time t0 is determined to be B, and time t0 is taken as the effective time of training action B. Then, several adjacent effective times are collected to obtain the effective time period of training action B.

[0018] Step 24: When the formula y = 1 - e -x For x > 0, y is a function that takes values ​​from 0 to 1, and y increases as x increases. Then, when |V n -W n When the deviation value increases, it indicates that there is a large difference between the nth vector in the reference vector set and the nth vector in the actual vector set. If there are multiple vectors in the reference vector set that have large differences with their corresponding vectors in the actual vector set, then the deviation value will be larger, indicating that the training action is not in compliance. If the deviation value is small, it indicates that the training action is in compliance. Therefore, in this scheme, when the deviation value is less than the preset deviation threshold, the user's training action at time t0 is determined to be B.

[0019] Preferably, step 3, which yields the target duration for each training exercise, includes:

[0020] Step 31: Obtain the minimum window duration D of the time window. min and maximum window duration D max The minimum window duration D is obtained. min and maximum window duration D max Given a window with a duration of D0, extract several time windows within the valid time period based on the window duration D0.

[0021] Step 32: Obtain the time domain function corresponding to a certain time window win, subtract the minimum amplitude from the maximum amplitude in the time domain function to obtain the amplitude difference of the time window win, randomly extract several moments within the time window win, calculate the variance between the amplitudes corresponding to all moments, and obtain the target variance of the time window win.

[0022] Step 33: Obtain the amplitude difference and target variance for each time window, calculate the average of the amplitude difference and target variance respectively, and obtain the average amplitude difference F0 and average target variance S0 corresponding to the window duration D0. Obtain the duration characteristic value X = F0 * S0 for the window duration D0; calculate the minimum window duration D. min and maximum window duration D max The duration feature values ​​of several windows are used, and the window duration corresponding to the smallest duration feature value is taken as the target duration of training action B; thus, the target duration of each training action is obtained.

[0023] Preferably, step 4 involves promptly issuing early warnings to relevant personnel, including:

[0024] Step 41: Collect the training video during the current training session, obtain the current training action based on the reference vector set; obtain the target duration of the current training action, extract all time windows according to the target duration, generate the frequency domain function for each time window, calculate the average power frequency of each frequency domain function, and establish the first set.

[0025] Step 42: The average power frequency is MPF, and its calculation process is based on existing technology; the frequency domain feature describes the frequency distribution law of the signal and reflects the energy distribution of muscle activity in the frequency dimension. For each time period, a corresponding frequency domain function can be generated, which is a function of power changing with frequency. It monitors the fatigue trend after continuous archery. Therefore, if the difference between the first set and the second set is large, that is, the cosine similarity is small, it indicates that there is an abnormality in the user's current training action, and a timely warning is required.

[0026] Step 43: Obtain the effective time period of the current training action, extract the time window within the effective time period according to the target duration, generate the frequency domain function of each time window, calculate the average power frequency of each frequency domain function, and establish a second set; the number of elements in the first set and the second set are the same, calculate the cosine similarity between the first set and the second set, if the cosine similarity is less than the preset similarity threshold, it is determined that there is an abnormality in the training action, and timely warning is given to the relevant instructors.

[0027] A real-time muscle activity analysis system based on electromyography (EMG) signals includes an EMG data extraction module, an effective time period extraction module, a target duration calculation module, and an early warning module.

[0028] Electromyography (EMG) data extraction module: used to retrieve historical signal acquisition records of user training projects, extract the sampling frequency corresponding to the signal acquisition records, and obtain the target sampling frequency based on the server resource usage during signal acquisition; guide users to complete voluntary muscle contraction activities, and continuously acquire EMG data of the entire contraction process according to the target sampling frequency;

[0029] Effective time period extraction module: used to decompose the training program into several training movements, retrieve the baseline movement videos of various training movements; obtain the baseline vector set of each limb based on the position of adjacent limbs in the baseline movement video; and obtain the effective time period of each training movement based on the baseline vector set and the monitoring video of the corresponding time of electromyography data.

[0030] Target duration calculation module: used to obtain the minimum and maximum window durations of the time window, extract several window durations within the minimum and maximum window durations, divide the effective time period into several time windows, establish the time domain function corresponding to the effective time period, obtain the duration feature value of each window duration based on the time domain function, and obtain the target duration of each training action based on the duration feature value.

[0031] Early warning module: It is used to obtain the training actions during the current training, obtain each time window according to the target duration of the training actions, evaluate the abnormal situation of the training actions based on the frequency domain information within the time window, and promptly issue early warnings to relevant instructors.

[0032] Preferably, the electromyography (EMG) data extraction module includes a target sampling frequency acquisition unit and an EMG data extraction unit;

[0033] Target sampling frequency acquisition unit: used to retrieve historical signal acquisition records of user project training, extract the sampling frequency in each signal acquisition record; obtain the minimum sampling frequency preset by the system; and, based on the server's resource occupancy value during signal acquisition, use the sampling frequency when the resource occupancy value equals the preset available resource occupancy value as the maximum sampling frequency; and randomly obtain either the minimum sampling frequency or the maximum sampling frequency as the target sampling frequency.

[0034] Electromyography (EMG) data extraction unit: Used to guide users in completing voluntary muscle contraction activities, continuously collecting EMG data throughout the entire contraction process according to the target sampling frequency.

[0035] Preferably, the early warning module includes a first set establishment unit and an early warning unit;

[0036] The first set establishment unit is used to obtain the current training action; obtain the target duration of the current training action, extract all current time windows, generate the frequency domain function of each time window, calculate the average power frequency of each frequency domain function, and establish the first set.

[0037] Early warning unit: used to establish a second set based on the effective time period; calculate the cosine similarity between the first set and the second set, determine whether there are any abnormalities in the training actions, and promptly issue early warnings to relevant instructors.

[0038] Compared with the prior art, the beneficial effects of the present invention are:

[0039] 1. This invention provides a real-time muscle activity analysis system and method based on electromyography (EMG) signals, comprising: retrieving signal acquisition records of user training projects, extracting the corresponding sampling frequencies, and obtaining the target sampling frequency; acquiring EMG data of the user completing muscle contractions; decomposing the training project into several training movements, obtaining the reference vector set of each limb, and obtaining the effective time period of each training movement; dividing the effective time period into several time windows, establishing the time domain function corresponding to the effective time period, and obtaining the target duration of each training movement; acquiring the training movements during the current training, evaluating abnormal situations in the training movements, and providing early warning prompts to relevant instructors.

[0040] 2. This invention analyzes historical signal acquisition records combined with multi-dimensional features to monitor abnormal behaviors during the training process in real time, accurately identify abnormal training movements, and promptly provide early warnings to instructors, meeting the needs of real-time detection and feedback, and helping to reduce the risk of muscle injury for users. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating a real-time muscle activity analysis method based on electromyographic signals according to the present invention. Detailed Implementation

[0042] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. It should be noted that the terms "front," "rear," "left," "right," "up," and "down" used in the following description refer to directions in the accompanying drawings, and the terms "inner" and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0043] like Figure 1 As shown, this embodiment provides a technical solution for real-time muscle activity analysis based on electromyography signals, including the following steps:

[0044] (1) Retrieve the historical signal acquisition records of the user's project training, extract the sampling frequency corresponding to the signal acquisition records, and obtain the target sampling frequency based on the server's resource usage during the signal acquisition process; guide the user to complete the muscle voluntary contraction activity, and continuously collect electromyographic data of the entire contraction process based on the target sampling frequency.

[0045] Retrieve historical signal acquisition records from user project training and extract the sampling frequency from each signal acquisition record; obtain the system's preset minimum sampling frequency; and, based on the server's resource usage during signal acquisition, use the sampling frequency at which the resource usage equals the preset available resource usage as the maximum sampling frequency.

[0046] The target sampling frequency is randomly selected from the lowest and highest sampling frequencies; the user is guided to complete the voluntary muscle contraction activity, and electromyographic data of the entire contraction process are continuously collected according to the target sampling frequency.

[0047] (2) Decompose the training program into several training movements and retrieve the benchmark movement videos of various training movements; based on the position of adjacent limbs in the benchmark movement videos, obtain the benchmark vector set of each limb; based on the benchmark vector set and the monitoring video of the corresponding time of electromyography data, obtain the effective time period of each training movement.

[0048] The training program is broken down into several training movements, and the baseline movement videos of various training movements are retrieved. A certain time t is obtained from the baseline movement video corresponding to a certain training movement B, a three-dimensional model is established, and the position of each limb at time t is obtained. The adjacent limbs of a certain limb Z0 are taken as Z1, and the positions of the head and tail of limb Z0 at time t are taken as Ps0 and Pw0 respectively, and the positions of the head and tail of limb Z1 at time t are taken as Ps1 and Pw1 respectively.

[0049] In this embodiment, taking archery as the training project, and Z0 as the forearm (the limb between the wrist and elbow joints) and Z1 as the upper arm (the limb between the elbow and shoulder joints) as an example, we can analyze the following: let the beginning and end positions of the forearm be the wrist and elbow joints, and the beginning and end positions of the upper arm be the elbow joint and shoulder joints, respectively. That is, Ps0 is the wrist position, Pw0 is the elbow joint position, Ps1 is the elbow joint position, and Pw1 is the shoulder joint position. Since the forearm and upper arm are in different positions during training, it indicates that different training is being performed. Therefore, we can obtain a set of reference vectors, analyze and judge them, and obtain the corresponding training actions.

[0050] Set the number of limb segments to N. Obtain N position points within positions Ps0 and Pw0 at uniform length intervals, where N>2. The position points include Ps0 and Pw0, and are sorted from beginning to end. Similarly, obtain N position points within positions Ps1 and Pw1 at uniform length intervals, where the position points include Ps1 and Pw1, and also sort them from beginning to end. Obtain two position points with the same index in limbs Z0 and Z1, and use the vector pointing from the position point in limb Z0 to the position point in limb Z1 as a reference vector for limb Z0, thus obtaining the set of reference vectors for limb Z0.

[0051] Obtain the time corresponding to the electromyography (EMG) data and retrieve the corresponding monitoring video; obtain the positions of limbs Z0 and Z1 at a certain time t0 in the monitoring video, obtain the actual vector set corresponding to limb Z0, the number of elements in the actual vector set is N, and obtain the deviation value: Where e is a natural constant, Vn is the nth vector in the reference vector set, and Wn is the nth vector in the actual vector set; if the deviation value is less than the preset deviation threshold, the user's training action at time t0 is determined to be B, and time t0 is taken as the effective time of training action B. Then, several adjacent effective times are collected to obtain the effective time period of training action B.

[0052] (3) Obtain the minimum window duration and maximum window duration of the time window, extract several window durations within the minimum window duration and maximum window duration, divide the effective time period into several time windows, establish the time domain function corresponding to the effective time period, obtain the duration feature value of each window duration according to the time domain function, and obtain the target duration of each training action according to the duration feature value.

[0053] Obtain the minimum window duration Dmin and the maximum window duration Dmax of the time window, obtain a certain window duration D0 within the minimum window duration Dmin and the maximum window duration Dmax, and extract several time windows within the effective time period according to the window duration D0;

[0054] Obtain the time-domain function corresponding to a certain time window win, subtract the minimum amplitude from the maximum amplitude in the time-domain function to obtain the amplitude difference of the time window win, randomly extract several moments within the time window win, calculate the variance between the amplitudes corresponding to all moments, and obtain the target variance of the time window win;

[0055] Obtain the amplitude difference and target variance for each time window, and calculate the average of the amplitude difference and target variance to obtain the average amplitude difference F0 and average target variance S0 corresponding to the window duration D0. Obtain the duration feature value X = F0 * S0 for the window duration D0. Calculate the duration feature values ​​of several window durations within the minimum window duration Dmin and the maximum window duration Dmax, and take the window duration corresponding to the smallest duration feature value as the target duration of training action B. Then obtain the target duration of each training action.

[0056] In electromyography (EMG) signal analysis, such as sEMG or other physiological signal analysis, the time domain and frequency domain are the core analytical dimensions, describing the signal or muscle state from the perspectives of temporal variation and frequency distribution, respectively. Time domain features describe the amplitude changes of the signal over time, reflecting the strength, duration, and trend of muscle activity over time. Amplitude difference characterizes the differences between amplitude values ​​within a time window; the smaller the amplitude difference, the smaller the difference, and the more reliable the calculations below. Target variance characterizes the differences between amplitude distributions within a time window; the smaller the target variance, the smaller the difference, and the more reliable the calculations below. Therefore, in this scheme, duration feature values ​​can be used for judgment and analysis to obtain a reliable target duration, and the following example is given:

[0057] In this scheme, the minimum window duration Dmin and the maximum window duration Dmax are pre-set. For example, the minimum window duration Dmin is 0.3 seconds, and the maximum window duration Dmax is 1 second. The window duration D0 is extracted, satisfying: Dmin ≤ D0 ≤ Dmax. Taking D0 as 0.5 seconds as an example, the effective time period is divided into several time windows according to the window duration of 0.5 seconds. Then, the time-domain function corresponding to a certain time window win is obtained. Taking a maximum amplitude of 3 and a minimum amplitude of 1 as an example, the unit is... Given μV, the amplitude difference of time window win is obtained as 2. Then, several moments within time window win are obtained, and the target variance of time window win is obtained, for example, as 5. Then, the amplitude difference and target variance of each time window are obtained, and the average amplitude difference F0 and average target variance S0 are obtained. In this embodiment, the average amplitude difference F0 is 1.5 and the average target variance S0 is 3. The duration feature value X of window duration D0 is obtained as 4.5. Then, the duration feature values ​​of several window durations are obtained, and thus a reliable target duration is obtained.

[0058] (4) Obtain the training movements during the current training, obtain each time window based on the target duration of the training movements, evaluate the abnormal situation of the training movements based on the frequency domain information within the time window, and promptly issue warnings to relevant instructors.

[0059] Collect the training video during the current training session, obtain the current training action based on the reference vector set, obtain the target duration of the current training action, extract all time windows according to the target duration, generate the frequency domain function for each time window, calculate the average power frequency of each frequency domain function, and establish the first set.

[0060] The average power frequency is MPF, and its calculation process is based on existing technology. Frequency domain features describe the frequency distribution law of the signal and reflect the energy distribution of muscle activity in the frequency dimension. For each time period, a corresponding frequency domain function can be generated, which is a function of power changing with frequency. It monitors the fatigue trend after continuous archery. Therefore, if the difference between the first set and the second set is large, that is, the cosine similarity is small, it indicates that there is an abnormality in the user's current training action, and a timely warning is required.

[0061] The effective time period of the current training action is obtained. The time window within the effective time period is extracted according to the target duration, and the frequency domain function of each time window is generated. The average power frequency of each frequency domain function is calculated, and a second set is established. The number of elements in the first set and the second set are the same. The cosine similarity between the first set and the second set is calculated. If the cosine similarity is less than the preset similarity threshold, it is determined that there is an abnormality in the training action, and the relevant instructors are promptly alerted.

[0062] This embodiment also provides a real-time muscle activity analysis system based on electromyography (EMG) signals, including an EMG data extraction module, an effective time period extraction module, a target duration calculation module, and an early warning module. The EMG data extraction module includes a target sampling frequency acquisition unit and an EMG data extraction unit. The early warning module includes a first set establishment unit and an early warning unit. When the system executes the computer program, it implements the above-mentioned real-time muscle activity analysis method based on EMG signals. Since this real-time muscle activity analysis method based on EMG signals has been described in detail above, it will not be repeated here.

[0063] The technical means disclosed in this invention are not limited to those disclosed in the above embodiments, but also include technical solutions composed of any combination of the above technical features.

Claims

1. A method for real-time analysis of muscle activity based on electromyographic signals, characterized in that, Includes the following steps: Step 1: Retrieve the historical signal acquisition records of the user project training, extract the sampling frequency corresponding to the signal acquisition records, and obtain the target sampling frequency based on the server's resource usage during the signal acquisition process; It guides users to complete voluntary muscle contraction activities and continuously collects electromyographic data throughout the entire contraction process based on the target sampling frequency; Step 2: Decompose the training program into several training movements and retrieve the baseline movement videos for each training movement; based on the positional relationship between adjacent limbs in the baseline movement videos, obtain the baseline vector set for each limb; based on the baseline vector set and the monitoring video corresponding to the electromyography data at the corresponding time, obtain the effective time period for each training movement. Step 3: Obtain the minimum and maximum window durations of the time window, extract several window durations within the minimum and maximum window durations, divide the effective time period into several time windows, establish a time domain function corresponding to the effective time period, obtain the duration feature value of each window duration based on the time domain function, and obtain the target duration of each training action based on the duration feature value; Step 3, obtaining the target duration of each training action, specifically includes the following steps: Step 31: Obtain the minimum window duration D of the time window. min and maximum window duration D max The minimum window duration D is obtained. min and maximum window duration D max Given a window with a duration of D0, extract several time windows within the valid time period based on the window duration D0. Step 32: Obtain the time domain function corresponding to a certain time window win, subtract the minimum amplitude from the maximum amplitude in the time domain function to obtain the amplitude difference of the time window win, randomly extract several moments within the time window win, calculate the variance between the amplitudes corresponding to all moments, and obtain the target variance of the time window win. Step 33: Obtain the amplitude difference and target variance for each time window, calculate the average of the amplitude difference and target variance respectively, and obtain the average amplitude difference F0 and average target variance S0 corresponding to the window duration D0. Obtain the duration characteristic value X = F0 * S0 for the window duration D0; calculate the minimum window duration D. min and maximum window duration D max The duration feature values ​​of several windows are used, and the window duration corresponding to the smallest duration feature value is taken as the target duration of training action B; thus, the target duration of each training action is obtained. Step 4: Obtain the training movements during the current training session. Based on the target duration of the training movements, obtain each time window. Based on the frequency domain information within the time windows, assess any abnormalities in the training movements and promptly issue warnings to relevant instructors. Step 4, promptly issuing warnings to relevant instructors, includes the following steps: Step 41: Collect the training video during the current training session, obtain the current training action based on the reference vector set; obtain the target duration of the current training action, extract all time windows according to the target duration, generate the frequency domain function for each time window, calculate the average power frequency of each frequency domain function, and establish the first set. Step 42: Obtain the effective time period of the current training action, extract the time window within the effective time period according to the target duration, generate the frequency domain function of each time window, calculate the average power frequency of each frequency domain function, and establish a second set; the number of elements in the first set and the second set are the same, calculate the cosine similarity between the first set and the second set, if the cosine similarity is less than the preset similarity threshold, it is determined that there is an abnormality in the training action, and timely warning is given to the relevant instructors.

2. The method for real-time analysis of muscle activity based on electromyographic signals according to claim 1, characterized in that, Step 1, continuously acquiring electromyographic data throughout the entire contraction process, includes: retrieving historical signal acquisition records from the user's training program and extracting the sampling frequency from each signal acquisition record; obtaining the system's preset minimum sampling frequency; based on the server's resource usage during signal acquisition, selecting the sampling frequency at which the resource usage equals the preset available resource usage as the maximum sampling frequency; randomly obtaining either the minimum or maximum sampling frequency as the target sampling frequency; guiding the user to complete voluntary muscle contraction activities and continuously acquiring electromyographic data throughout the entire contraction process according to the target sampling frequency.

3. The method for real-time analysis of muscle activity based on electromyographic signals according to claim 1, characterized in that, Step 2, obtaining the effective time period for each training exercise, includes the following steps: Step 21: Decompose the training program into several training movements and retrieve the baseline movement videos for each training movement; obtain a time t from the baseline movement video corresponding to a certain training movement B, establish a 3D model, and obtain the position of each limb at time t. Take the adjacent limbs of a certain limb Z0 as Z1, and take the positions of the head and tail of limb Z0 at time t as P respectively. s 0 and P w 0 The positions of the head and tail of limb Z1 at time t are respectively designated as P. s 1 and P w 1 ; Step 22: Set the number of limb segments to N, and obtain the positions P according to uniform length intervals. s 0 and P w 0 There are N points within a given area, where N > 2, and the points include P. s 0 and P w 0 And sort the positions in order from beginning to end. Similarly, obtain the positions P according to uniform length intervals. s 1 and P w 1 There are N location points within the range, including point P. s 1 and P w 1 The positions are also sorted in order from beginning to end; two positions with the same index are obtained in limb Z0 and limb Z1, and the vector pointing from the position point in limb Z0 to the position point in limb Z1 is used as a reference vector of limb Z0, thus obtaining the reference vector set of limb Z0. Step 23: Obtain the time corresponding to the electromyography data and retrieve the corresponding monitoring video; obtain the positions of limbs Z0 and Z1 at a certain time t0 in the monitoring video, obtain the actual vector set corresponding to limb Z0, the number of elements in the actual vector set is N, and obtain the deviation value: Where e is the natural constant, and V n W is the nth vector in the reference vector set. n It is the nth vector in the actual vector set; If the deviation value is less than the preset deviation threshold, the user's training action at time t0 is determined to be B. Time t0 is taken as the effective time of training action B. Then, several adjacent effective times are collected to obtain the effective time period of training action B.

4. A real-time muscle activity analysis system, used to execute the real-time muscle activity analysis method based on electromyographic signals as described in any one of claims 1-3, characterized in that, The system includes an electromyography data extraction module, an effective time period extraction module, a target duration calculation module, and an early warning module. Electromyography (EMG) data extraction module: used to retrieve historical signal acquisition records of user project training, extract the sampling frequency corresponding to the signal acquisition records, and obtain the target sampling frequency based on the server's resource usage during signal acquisition; It guides users to complete voluntary muscle contraction activities and continuously collects electromyographic data throughout the entire contraction process based on the target sampling frequency; Effective time period extraction module: used to break down training items into several training movements and retrieve the baseline movement videos of various training movements; Based on the positional information between adjacent limbs in the baseline movement video, a set of baseline vectors for each limb is obtained; based on the set of baseline vectors and the monitoring video corresponding to the electromyography data at the time, the effective time period for each training movement is obtained. Target duration calculation module: used to obtain the minimum and maximum window durations of the time window, extract several window durations within the minimum and maximum window durations, divide the effective time period into several time windows, establish the time domain function corresponding to the effective time period, obtain the duration feature value of each window duration based on the time domain function, and obtain the target duration of each training action based on the duration feature value. Early warning module: It is used to obtain the training actions during the current training, obtain each time window according to the target duration of the training actions, evaluate the abnormal situation of the training actions based on the frequency domain information within the time window, and promptly issue early warnings to relevant instructors.

5. The real-time muscle activity analysis system according to claim 4, characterized in that, The electromyography data extraction module includes a target sampling frequency acquisition unit and an electromyography data extraction unit; Target sampling frequency acquisition unit: used to retrieve historical signal acquisition records of user project training and extract the sampling frequency from each signal acquisition record; Obtain the system's preset minimum sampling frequency. Based on the server's resource usage value during signal acquisition, use the sampling frequency at which the resource usage value equals the preset available resource usage value as the maximum sampling frequency. Randomly obtain either the minimum or the maximum sampling frequency as the target sampling frequency. Electromyography (EMG) data extraction unit: Used to guide users in completing voluntary muscle contraction activities, continuously collecting EMG data throughout the entire contraction process according to the target sampling frequency.

6. The real-time muscle activity analysis system according to claim 4, characterized in that, The early warning module includes a first set establishment unit and an early warning unit; The first set establishment unit is used to obtain the current training action; obtain the target duration of the current training action, extract all current time windows, generate the frequency domain function of each time window, calculate the average power frequency of each frequency domain function, and establish the first set. Early warning unit: used to establish a second set based on the effective time period; Calculate the cosine similarity between the first set and the second set to determine if there are any abnormalities in the training actions, and promptly issue warnings to relevant instructors.

Images