A motion information analysis system and method based on multi-channel surface electromyography signal acquisition
Through a multi-channel surface electromyography signal acquisition system, combined with filtering, segmentation window function, principal component analysis and genetic algorithm optimization of support vector machine, the problem of high cost and complex data analysis in existing technologies is solved, and low-cost and accurate fitness movement recognition and analysis are achieved, which improves the analysis accuracy of user movement posture and the human-computer interaction experience.
Patent Information
- Application Number
- CN202211345611.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-10-31
- Publication Date
- 2025-09-19
- Estimated Expiration
- 2042-10-31
AI Technical Summary
In the existing technology, fitness movement recognition systems based on computer vision are expensive and the data analysis is complex, and there is a lack of effective intelligent fitness movement recognition and analysis systems based on sEMG signals.
A multi-channel surface electromyography signal acquisition system is used, combined with filtering, segmentation window function, principal component analysis and genetic algorithm to optimize support vector machine, and decision fusion is performed through DS evidence theory to achieve recognition and analysis of fitness movements.
It achieves low-cost and accurate fitness movement recognition, improves the analysis accuracy of user movement posture and human-computer interaction experience, and is suitable for ordinary fitness enthusiasts.
Smart Images
Figure CN115712862B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the fields of communication technology, detection and identification, and computer technology, and is a software and hardware integrated system for data acquisition and data processing, and in particular to a motion information acquisition and analysis system based on multi-channel surface electromyography signal acquisition, transmission, and intelligent analysis, and a control method thereof. Background Art
[0002] With advancements in technology and the development of artificial intelligence, fitness is becoming increasingly intelligent, and smart fitness is, to a certain extent, driving the implementation of national fitness programs. Intelligent detection technology is widely used in wearable fitness systems, such as smart bracelets and intelligent detection systems. However, research on fitness movement recognition based on sEMG signals is limited, and there is a need to design intelligent analysis systems specifically for fitness movement recognition and analysis.
[0003] Existing technologies primarily rely on computer vision algorithms to assess and analyze human motion posture, encompassing target detection, motion target classification, human tracking, and behavioral understanding and description. However, implementing computer vision requires high costs and a large amount of data for analysis and optimization. sEMG bioelectrical signals, recorded from the skin surface during neuromuscular activity, contain a wealth of information related to human motion. Their time-frequency and nonlinear dynamic characteristics are sensitive to the system's activity level and state, making them an effective method for evaluating neuromuscular function.
[0004] Therefore, in view of the technical defects of the existing technology, it is necessary to propose a solution to solve the technical problems of the existing technology. Summary of the Invention
[0005] The purpose of this invention is to propose a motion information analysis system based on multi-channel surface electromyography signal acquisition, which can realize the recognition and analysis of fitness movements by recording multi-channel surface electromyography signals during the user's exercise process, thereby improving the user's exercise posture.
[0006] In order to solve the technical problems existing in the prior art, the technical solutions of the present invention are as follows:
[0007] A motion information analysis system based on multi-channel surface electromyography signal acquisition is characterized by comprising at least a data acquisition module and a data processing module, wherein the data acquisition module is used to collect at least 8 channels of surface electromyography signals during a user's motion and send the data to the data processing module after data processing, and the data processing module is used to process the collected electromyography data and output an evaluation result, and at least perform the following steps:
[0008] Step S1: Use the filter function to filter the data and use the segmentation window function to segment the original data. Note the number of sampling points N, the number of samples n, and the sample value x. n , the sample mean is recorded as u, and the sample standard deviation is recorded as σ. The sampling points at different positions are k, the corresponding frequency spectrum value is P(k), and the frequency amplitude of the corresponding point is f k , and then calculate the root mean square value RMS, skewness Skew, centroid frequency S1, frequency root mean square S2, frequency standard deviation S3 and other features of the extracted segmented data to form a feature vector group. The calculation formula is as follows:
[0009]
[0010]
[0011]
[0012]
[0013]
[0014] Step S2: Principal component analysis is used to analyze the principal component contribution rate and cumulative contribution rate of each input feature variable, select the feature vectors with the largest contribution rate and reconstruct a new vector group to achieve feature dimensionality reduction. Then, a genetic algorithm is used to optimize the parameters of the kernel function in the SVM. The penalty factor C and kernel parameter g of different kernel functions are optimized separately, including the linear kernel function LKF, the polynomial kernel function PKF, the Gaussian kernel function RBF, and the Sigmoid kernel function. The calculation formula is as follows:
[0015] LKF=x t z
[0016] PKF=(gx t z+c) d
[0017] RBF=exp(-g||xz|| 2 )
[0018] Sigmod=tanh(gx t z+c)
[0019] Among them, x and z are n-dimensional input values, x t represents the transpose of x, c represents the offset, and g is the kernel parameter. The kernel parameter g of different kernel functions is set as follows:
[0020] LKF has no special parameters that need to be set;
[0021] PKF has three parameters. d is used to set the highest order of PKF, which is d in the formula, and the default value is 3. g is used to set the gamma parameter in the kernel function, and the default value is 1 / k (k is the number of categories). c is used to set coef0 in the kernel function, and the default value is 0.
[0022] RBF has one parameter. g is used to set the gamma parameter in the kernel function, and the default value is 1 / k;
[0023] Sigmoid has two parameters. g is used to set the gamma parameter in the kernel function, and the default value is 1 / k. -r is used to set coef0 in the kernel function, and the default value is 0;
[0024] In order to make up for the problem of insufficient intervals between some linearly inseparable sample points, we introduce a relaxation factor ζ(i) for each sample point so that the final interval is greater than or equal to 1. Note: w=(w1,w2,...,w d ) T Normal vector, b is the displacement. ||w|| represents the binorm of w. Find the sum of the squares of all elements and then take the square root. The larger the C value, the more it penalizes the slack variable. The optimization goal is to make the slack factor close to 0, that is, the penalty for misclassification becomes larger. At this time, the accuracy is high when testing the training set, but the generalization ability is weak. The smaller the C value, the smaller the penalty for misclassification, allowing for error tolerance, treating them as noise points, and having stronger generalization ability. After adding the relaxation factor ζ(i), the objective function and constraints of the support vector machine optimization are:
[0025]
[0026] sty (i) (w T Φ(x (i) )+b)≥1-ζ(i),i=1,2,...n
[0027] Step S3: Fuse the output results of the four different kernel functions, adopt DS evidence theory, set the recognition framework of this algorithm to Θ, and the basic probability distribution on this recognition framework is P(Θ)→[0,1]. Set S to be the entire event set, that is, the training input label parameter in this algorithm, then for Set A as an event set, B and C as small events in the set, To normalize the constant, two independent bodies of evidence m1 and m2 can be combined using Dempster's rule, as follows:
[0028]
[0029] In the decision-making process, a simple method to convert the belief function into a probability function can be used, namely the Pignistic probability transformation, and the calculation formula is:
[0030]
[0031] Finally, the performance of the optimized SVM is significantly improved.
[0032] By adopting the technical solution of the present invention, the multi-channel surface electromyography signals of the user during exercise are recorded to realize the recognition and analysis of fitness movements, thereby improving the user's exercise posture.
[0033] As a further improvement, a host computer software terminal is provided, in which a data processing module is provided, which is wirelessly connected to the data acquisition module for real-time display of the evaluation results.
[0034] As a further improvement scheme, the data acquisition module includes at least a sensor module, an amplification and filtering circuit, a Bluetooth module and a digital-to-analog conversion module, wherein the sensor module is used to collect at least 8 channels of surface electromyography signals during the user's movement; the amplification and filtering circuit is constructed by AD623 and TLC2274 to realize signal preprocessing; the digital-to-analog conversion module is used to convert analog signals into digital signals; and the Bluetooth module is used to transmit data to the host computer.
[0035] As a further improvement, the amplification and filtering circuit includes a differential input precision amplifier AD623 and a rail-to-rail high-speed operational amplifier TLC2274, wherein the AD623 is used to provide power supply output amplitude gain, and the TLC2274 is used to provide a larger output range, providing a basis for outputting a larger gain signal.
[0036] As a further improvement, the Bluetooth module is used to transmit data to a host computer, transmitting at least 8 channels of surface electromyography signals.
[0037] As a further improvement, the data acquisition module uses Arduino MEGA 2560 as the control chip.
[0038] As a further improvement solution, the host computer software terminal is a smart computer.
[0039] As a preferred technical solution, the amplifying and filtering circuit includes a differential input precision amplifier AD623 and a rail-to-rail high-speed operational amplifier TLC2274, which are used to provide power supply output amplitude gain and a larger output range.
[0040] As a preferred technical solution, the Bluetooth module adopts wireless transmission to transmit at least 8 channels of surface electromyography signals.
[0041] In order to solve the technical problems existing in the prior art, the present invention proposes a motion information analysis method based on multi-channel surface electromyography signal acquisition, which at least includes the following steps:
[0042] Step S1: Use the filter function to filter the data and use the segmentation window function to segment the original data. Note the number of sampling points N, the number of samples n, and the sample value x. n , the sample mean is recorded as u, and the sample standard deviation is recorded as σ. The sampling points at different positions are k, the corresponding frequency spectrum value is P(k), and the frequency amplitude of the corresponding point is f k , and then calculate the root mean square value RMS, skewness Skew, centroid frequency S1, frequency root mean square S2, frequency standard deviation S3 and other features of the extracted segmented data to form a feature vector group. The calculation formula is as follows:
[0043]
[0044]
[0045]
[0046]
[0047]
[0048] Step S2: Principal component analysis is used to analyze the principal component contribution rate and cumulative contribution rate of each input feature variable, select the feature vectors with the largest contribution rate and reconstruct a new vector group to achieve feature dimensionality reduction. Then, a genetic algorithm is used to optimize the parameters of the kernel function in the SVM. The penalty factor C and kernel parameter g of different kernel functions are optimized separately, including the linear kernel function LKF, the polynomial kernel function PKF, the Gaussian kernel function RBF, and the Sigmoid kernel function. The calculation formula is as follows:
[0049] LKF=x t z
[0050] PKF=(gx t z+c) d
[0051] RBF=exp(-g||xz|| 2 )
[0052] Sigmod=tanh(gx t z+c)
[0053] Where x and z are n-dimensional input values, xt represents the transpose of x, c represents the offset, and g is the kernel parameter. The kernel parameter g for different kernel functions is set as follows:
[0054] LKF has no special parameters that need to be set;
[0055] PKF has three parameters. d is used to set the highest order of PKF, which is d in the formula, and the default value is 3. g is used to set the gamma parameter in the kernel function, and the default value is 1 / k (k is the number of categories). c is used to set coef0 in the kernel function, and the default value is 0.
[0056] RBF has one parameter. g is used to set the gamma parameter in the kernel function, and the default value is 1 / k;
[0057] Sigmoid has two parameters. g is used to set the gamma parameter in the kernel function, and the default value is 1 / k. -r is used to set coef0 in the kernel function, and the default value is 0;
[0058] In order to make up for the problem of insufficient intervals between some linearly inseparable sample points, we introduce a relaxation factor ζ(i) for each sample point so that the final interval is greater than or equal to 1. Note: w = (w1, w2, ..., w d ) T Normal vector, b is the displacement. ||w|| represents the binorm of w. Find the sum of the squares of all elements and then take the square root. The larger the C value, the more it penalizes the slack variable. The optimization goal is to make the slack factor close to 0, that is, the penalty for misclassification becomes larger. At this time, the accuracy is high when testing the training set, but the generalization ability is weak. The smaller the C value, the smaller the penalty for misclassification, allowing for error tolerance, treating them as noise points, and having stronger generalization ability. After adding the relaxation factor ζ(i), the objective function and constraints of the support vector machine optimization are:
[0059]
[0060]
[0061] Step S3: Fuse the output results of the four different kernel functions, adopt DS evidence theory, set the recognition frame of this algorithm to Θ, and the basic probability distribution on this recognition frame is P(Θ)→[0,1]. Set S to be the entire event set, that is, the training input label parameter in this algorithm, then for Set A as an event set, B and C as small events in the set, To normalize the constant, two independent bodies of evidence m1 and m2 can be combined using Dempster's rule, as follows:
[0062]
[0063] In the decision-making process, a simple method to convert the belief function into a probability function can be used, namely the Pignistic probability transformation, and the calculation formula is:
[0064]
[0065] Finally, the performance of the optimized SVM is significantly improved.
[0066] Compared with the prior art, the present invention has the following technical effects:
[0067] 1. Compared with the human motion posture recognition technology solutions based on computer vision on the market, the present invention has low cost, is easy to carry, and is suitable for ordinary fitness enthusiasts.
[0068] 2. Compared with the mainstream method on the market that extracts muscle fatigue characteristics, heart rate and other parameters for motion posture recognition, the present invention extracts muscle surface EMG signals for processing and recognition, making the results more accurate.
[0069] 3. The present invention is different from sports competition equipment on the market. It can use APP to realize the visualization processing of electromyographic data and comprehensively analyze the user's exercise status to optimize the human-computer interaction experience when the user uses it. BRIEF DESCRIPTION OF THE DRAWINGS
[0070] Figure 1 It is a structural diagram of the intelligent motion information collection and analysis system of the present invention.
[0071] Figure 2 This is a software flow chart for the present invention to display evaluation results in real time.
[0072] Figure 3 This is a flow chart of the improved algorithm based on principal component analysis and genetic algorithm in the present invention.
[0073] The following specific implementation will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION
[0074] The technical solution provided by the present invention will be further described below with reference to the accompanying drawings.
[0075] Compared to the existing technology, the present invention provides a motion information acquisition and analysis system based on multi-channel surface electromyography signal acquisition, transmission, and intelligent analysis. This system collects and extracts eigenvalues from multi-channel sEMG signals during different fitness movements, displays the analysis results on a host computer, and provides reasonable fitness recommendations. By recording multi-channel surface electromyography signals during a user's exercise, fitness movements can be identified and analyzed, thereby improving the user's exercise posture. The fitness movement classification model is trained and tested based on an improved SVM algorithm based on principal component analysis and genetic algorithms. The classification performance is superior to that of a single kernel function, and various fitness movements can be more accurately identified.
[0076] See also Figure 1 , shown is a principle block diagram of an intelligent motion information acquisition and analysis system of the present invention, including a hardware module and a data processing module set in the host computer software, and the data acquisition module and the host computer communicate wirelessly;
[0077] The hardware collects sEMG signals from eight locations: the biceps, triceps, deltoids, and pectoralis major. This data is collected using a reference electrode and 16 six-lead electromyographic sensors. Disposable electrocardiogram (ECG) electrodes are preferred. The electromyographic signals from each muscle point are collected using a pair of positive and negative electrodes. An area without muscle activity serves as the reference electrode. After processing by the hardware's amplification and filtering circuits, the signals are fed into an Arduino MEGA 2560 module. The included digital-to-analog converter converts the analog signals into digital signals, which are then wirelessly transmitted to a host computer via Bluetooth. A suitable design ensures that the module meets the requirements as perfectly as possible. Therefore, we placed the data processing on the host computer. This approach minimizes the hardware system's power consumption, improves system operation time, and reduces hardware components, reducing size and weight, thus minimizing the user's fitness experience.
[0078] See also Figure 2 , which is a flowchart of the software used for real-time display of evaluation results in the present invention, at least includes the steps of serial port configuration initialization, data reception, data decoding, and motion information visualization, wherein:
[0079] The serial port configuration initialization interacts with the serial port hardware device through the web serial port API provided by the Chromium core, and pre-agreed on the serial port baud rate 115200 bit / s and other transmission parameters for communication;
[0080] The data receiving is received asynchronously via a single thread through the serial port and stored in a local cache, and the receiving time is added to the timestamp;
[0081] The data decoding process sorts the received binary data according to the pre-agreed communication format specification to distinguish the valid data of each channel;
[0082] The motion information visualization is based on the front-end framework Quasar of VoeJS to write the front-end page, and uses the Echarts data visualization framework and sampling method to display the received data in the form of dynamic charts.
[0083] See also Figure 3 , which is a flow chart of the improved algorithm based on principal component analysis and genetic algorithm in the present invention. The received sEMG signal is processed to complete the following algorithm:
[0084] Step S1: Use the filter function to filter the data and use the segmentation window function to segment the original data. Note the number of sampling points N, the number of samples n, and the sample value x. n , the sample mean is recorded as u, and the sample standard deviation is recorded as σ. The sampling points at different positions are k, the corresponding frequency spectrum value is P(k), and the frequency amplitude of the corresponding point is f k , and then calculate the root mean square value RMS, skewness Skew, centroid frequency S1, frequency root mean square S2, frequency standard deviation S3 and other features of the extracted segmented data to form a feature vector group. The calculation formula is as follows:
[0085]
[0086]
[0087]
[0088]
[0089]
[0090] Step S2: Principal component analysis is used to analyze the principal component contribution rate and cumulative contribution rate of each input feature variable, select the feature vectors with the largest contribution rate and reconstruct a new vector group to achieve feature dimensionality reduction. Then, a genetic algorithm is used to optimize the parameters of the kernel function in the SVM. The penalty factor C and kernel parameter g of different kernel functions are optimized separately, including the linear kernel function LKF, the polynomial kernel function PKF, the Gaussian kernel function RBF, and the Sigmoid kernel function. The calculation formula is as follows:
[0091] LKF=x t z
[0092] PKF=(gx t z+c) d
[0093] RBF=exp(-g||xz|| 2 )
[0094] Sigmod=tanh(gx t z+c)
[0095] Among them, x and z are n-dimensional input values, x t represents the transpose of x, c represents the offset, and g is the kernel parameter. The kernel parameter g of different kernel functions is set as follows:
[0096] LKF has no special parameters that need to be set;
[0097] PKF has three parameters. d is used to set the highest order of PKF, which is d in the formula, and the default value is 3. g is used to set the gamma parameter in the kernel function, and the default value is 1 / k (k is the number of categories). c is used to set coef0 in the kernel function, and the default value is 0.
[0098] RBF has one parameter. g is used to set the gamma parameter in the kernel function, and the default value is 1 / k;
[0099] Sigmoid has two parameters. g is used to set the gamma parameter in the kernel function, and the default value is 1 / k. -r is used to set coef0 in the kernel function, and the default value is 0;
[0100] In order to make up for the problem of insufficient intervals between some linearly inseparable sample points, we introduce a relaxation factor ζ(i) for each sample point so that the final interval is greater than or equal to 1. Note: w = (w1, w2, ..., w d ) T Normal vector, b is the displacement. ||w|| represents the binorm of w. Find the sum of the squares of all elements and then take the square root. The larger the C value, the more it penalizes the slack variable. The optimization goal is to make the slack factor close to 0, that is, the penalty for misclassification becomes larger. At this time, the accuracy is high when testing the training set, but the generalization ability is weak. The smaller the C value, the smaller the penalty for misclassification, allowing for error tolerance, treating them as noise points, and having stronger generalization ability. After adding the relaxation factor ζ(i), the objective function and constraints of the support vector machine optimization are:
[0101]
[0102] sty (i) (w T Φ(x (i) )+b≥1-ζ(i), i=1, 2,...n
[0103] Step S3: Fuse the output results of the four different kernel functions, adopt DS evidence theory, set the recognition frame of this algorithm to Θ, and the basic probability distribution on this recognition frame is P(Θ)→[0,1]. Set S to be the entire event set, that is, the training input label parameter in this algorithm, then for Set A as an event set, B and C as small events in the set, To normalize the constant, two independent bodies of evidence m1 and m2 can be combined using Dempster's rule, as follows:
[0104]
[0105] In the decision-making process, a simple method to convert the belief function into a probability function can be used, namely the Pignistic probability transformation, and the calculation formula is:
[0106]
[0107] Finally, the performance of the optimized SVM is significantly improved.
[0108] This invention utilizes the DS evidence fusion algorithm, resulting in an optimized SVM with significantly higher performance than a single kernel function, enabling more accurate recognition of various movements. The sEMG signal acquisition system, based on the interdisciplinary application of information processing and biomedicine within the fields of electronic information and communications engineering, holds great promise for future development.
[0109] The host computer uses software to visualize myoelectric data. After the user selects a training program, the software can display muscle force in real time, enabling efficient information exchange. This allows users to better adjust their training movements and provides more targeted instruction for coaches.
[0110] This technical solution realizes multi-channel surface electromyography signal acquisition, transmission and intelligent analysis of motion posture recognition; at the same time, the hardware system is lightweight and simple, and the collection of motion data is completed without affecting the user's exercise experience; during the fitness process, the user can know the motion posture data of the fitness process in real time from the software end. Through this data, the user can comprehensively evaluate the effect of this exercise and improve the user's exercise experience.
[0111] It is also worth pointing out that:
[0112] 1. The data acquisition module and the mobile phone can be connected not only by Bluetooth transmission, but also by wired connection methods such as USB data cable and other wireless connection technologies;
[0113] 2. The 8-channel sensor of the data acquisition module can also be expanded to other multi-channel sEMG acquisition sensors;
[0114] 3. The main control chip of the data acquisition module supports various control units on the market;
[0115] 4. This technical solution is based on principal component analysis and genetic algorithm improvement for posture recognition and is generally applicable to various fitness movements.
[0116] The above embodiments are only intended to help understand the method and core concept of the present invention. It should be noted that, without departing from the principles of the present invention, a number of improvements and modifications may be made to the present invention by those skilled in the art, and such improvements and modifications also fall within the scope of protection of the claims of the present invention.
[0117] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A motion information analysis system based on multi-channel surface electromyography signal acquisition, characterized in that: The system includes at least a data acquisition module and a data processing module, wherein the data acquisition module is used to collect at least 8 channels of surface electromyographic signals during the user's exercise and send them to the data processing module after data processing. The data processing module is used to process the collected electromyographic data and output an evaluation result, and at least perform the following steps: Step S1: Use the filter function to filter the data and use the segmentation window function to segment the original data. The number of sampling points is N, the number of samples is n, the sample value is xn, the sample mean is u, and the standard deviation of the sample is σ. The sampling points at different positions are k, the corresponding frequency spectrum value is P(k), and the frequency amplitude of the corresponding point is f k , and then calculate the root mean square value RMS, skewness Skew, centroid frequency S1, frequency root mean square S2, frequency standard deviation S3 and other features of the extracted segmented data to form a feature vector group. The calculation formula is as follows: Step S2: Principal component analysis is used to analyze the principal component contribution rate and cumulative contribution rate of each input feature variable, select the feature vector with the largest contribution rate and reconstruct a new vector group to achieve feature dimensionality reduction; then the genetic algorithm is used to optimize the parameters of the kernel function in the SVM, and the penalty factor C and kernel parameter g of different kernel functions are optimized respectively, including the linear kernel function LKF, the polynomial kernel function PKF, the Gaussian kernel function RBF and the Sigmoid kernel function. The calculation formula is as follows: LKF=x t z PKF=(gx t (z+c) d RBF=exp(-g‖x-z‖ 2 ) Sigmod=tanh(gx t z+c) Among them, x and z are n-dimensional input values, x t represents the transpose of x, c represents the offset, and g is the kernel parameter; the kernel parameter g of different kernel functions is set as follows: PKF has three parameters; d is used to set the highest order of PKF; g is used to set the gamma parameter in the kernel function; c is used to set coef0 in the kernel function; RBF has one parameter; g is used to set the gamma parameter in the kernel function; Sigmoid has two parameters; g is used to set the gamma parameter in the kernel function; -r is used to set coef0 in the kernel function; At the same time, a relaxation factor ζ(i) is introduced for each sample point so that the final interval is greater than or equal to 1; Note: w=(w1,w2,...,w d ) T Normal vector, b is the displacement; ||w|| represents the binorm of w. After adding the relaxation factor ζ(i), the objective function and constraints of the support vector machine optimization are: s.t.y (i) (w T Φ(x (i) )+b)≥1-ζ(i),i=1,2,...n Step S3: Fuse the output results of the four different kernel functions. According to the DS evidence theory, set the recognition frame of this algorithm to Θ, then the basic probability distribution on the recognition frame is P(Θ)→[0,1]; set S to the entire event set, that is, the training input label parameter in this algorithm, then for Set A as an event set, B and C as small events in the set, To normalize the constant, two independent bodies of evidence m1 and m2 can be combined using Dempster's rule, as follows: In the decision-making process, a method is used to convert the belief function into a probability function, namely the Pignistic probability transformation, and the calculation formula is: Ultimately, the performance of the optimized SVM is improved.
2. The motion information analysis system based on multi-channel surface electromyography signal acquisition according to claim 1 is characterized in that: A host computer software terminal is also provided, in which a data processing module is provided, which is connected to the data acquisition module in a wireless manner for real-time display of evaluation results.
3. The motion information analysis system based on multi-channel surface electromyography signal acquisition according to claim 2, characterized in that: The data acquisition module includes at least a sensor module, an amplifying and filtering circuit, a Bluetooth module and a digital-to-analog conversion module, wherein the sensor module is used to collect at least 8 channels of surface electromyography signals during the user's movement; the amplifying and filtering circuit is constructed by AD623 and TLC2274 to realize signal preprocessing; the digital-to-analog conversion module is used to convert analog signals into digital signals; and the Bluetooth module is used to transmit data to the host computer.
4. The motion information analysis system based on multi-channel surface electromyography signal acquisition according to claim 3 is characterized in that: The amplification and filtering circuit includes a differential input precision amplifier AD623 and a rail-to-rail high-speed operational amplifier TLC2274, wherein the AD623 is used to provide power supply output amplitude gain, and the TLC2274 is used to provide a larger output range, providing a basis for outputting a larger gain signal.
5. The motion information analysis system based on multi-channel surface electromyography signal acquisition according to claim 3 is characterized in that: The Bluetooth module is used to transmit data to a host computer, and transmits at least 8 channels of surface electromyography signals.
6. The motion information analysis system based on multi-channel surface electromyography signal acquisition according to claim 3, characterized in that: The data acquisition module uses Arduino MEGA 2560 as the control chip.
7. The motion information analysis system based on multi-channel surface electromyography signal acquisition according to claim 3 is characterized in that: The host computer software terminal is a smart computer.
8. A motion information analysis method based on multi-channel surface electromyography signal acquisition, characterized in that: At least the following steps are included: Step S1: Collect at least 8 channels of surface electromyographic signal data during the user's movement, filter the data using a filter function, and segment the original data using a segmentation window function. Record the number of sampling points N, the number of samples n, and the sample value x. n , the sample mean is recorded as u, the sample standard deviation is recorded as σ; the sampling points at different positions are k, the corresponding frequency spectrum value is P(k), and the frequency amplitude of the corresponding point is f k , and then calculate the root mean square value RMS, skewness Skew, centroid frequency S1, frequency root mean square S2, frequency standard deviation S3 and other features of the extracted segmented data to form a feature vector group. The calculation formula is as follows: Step S2: Principal component analysis is used to analyze the principal component contribution rate and cumulative contribution rate of each input feature variable, select the feature vector with the largest contribution rate and reconstruct a new vector group to achieve feature dimensionality reduction; then the genetic algorithm is used to optimize the parameters of the kernel function in the SVM, and the penalty factor C and kernel parameter g of different kernel functions are optimized respectively, including the linear kernel function LKF, the polynomial kernel function PKF, the Gaussian kernel function RBF and the Sigmoid kernel function. The calculation formula is as follows: LKF=x t z PKF=(gx t (z+c) d RBF=exp(-g‖x-z‖ 2 ) Sigmod=tanh(gx t z+c) Among them, x and z are n-dimensional input values, x t represents the transpose of x, c represents the offset, and g is the kernel parameter; the kernel parameter g of different kernel functions is set as follows: PKF has three parameters; d is used to set the highest order of PKF; g is used to set the gamma parameter in the kernel function; c is used to set coef0 in the kernel function; RBF has one parameter; g is used to set the gamma parameter in the kernel function; Sigmoid has two parameters; g is used to set the gamma parameter in the kernel function; -r is used to set coef0 in the kernel function; At the same time, a relaxation factor ζ(i) is introduced for each sample point so that the final interval is greater than or equal to 1; Note: w=(w1,w2,...,w d ) T Normal vector, b is the displacement; ||w|| represents the binorm of w. After adding the relaxation factor ζ(i), the objective function and constraints of the support vector machine optimization are: s.t.y (i) (w T Φ(x (i) )+b)≥1-ζ(i),i=1,2,...n Step S3: Fuse the output results of the four different kernel functions. According to the DS evidence theory, set the recognition frame of this algorithm to Θ, then the basic probability distribution on the recognition frame is P(Θ)→[0,1]; set S to the entire event set, that is, the training input label parameter in this algorithm, then for Set A as an event set, B and C as small events in the set, To normalize the constant, two independent bodies of evidence m1 and m2 can be combined using Dempster's rule, as follows: In the decision-making process, a method is used to convert the belief function into a probability function, namely the Pignistic probability transformation, and the calculation formula is: Ultimately, the performance of the optimized SVM is improved.