Electromyogram gait recognition and evaluation method and system based on machine learning

Through the electromyography gait recognition evaluation method based on machine learning, using sliding windows and diverse features to optimize the gait detection model, the problem of insufficient universality and accuracy of gait recognition in the existing technology is solved, and the problem of "unprecedented" is realized Efficient and real-time gait detection.

CN118902480APending Publication Date: 2024-11-08SHANDONG UNIV
View PDF 0 Cites 5 Cited by

Patent Information

Application Number
CN202410991420.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing gait recognition technology has poor universality, insufficient real-time and accuracy, and cannot effectively distinguish gaits of different ages and physical conditions. The sample characteristics are insufficient, which affects detection capabilities.

Method used

Using the EMG gait recognition evaluation method based on machine learning, the gait recognition evaluation method is used to obtain the human EMG signal, build a training set, extract and screen gait features, and build a machine learning model, using sliding windows and diversified time domain features such as mean squares. The gait detection model is optimized to improve real-time response and adaptability.

Benefits of technology

It realizes efficient, real-time and highly adaptable gait detection, which can accurately identify the current gait stage of pedestrians, adapt to different walking speeds and habits, and significantly improves the accuracy and universality of gait evaluation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118902480A_ABST
    Figure CN118902480A_ABST
Patent Text Reader

Abstract

The invention discloses an electromyogram gait recognition evaluation method and system based on machine learning, and the method comprises the steps: carrying out the feature extraction of human body electromyogram signals of a training set, and obtaining a plurality of gait features; performing feature screening from all gait features to obtain a plurality of most representative features; constructing a plurality of different feature combinations based on a plurality of most representative features; inputting each feature combination into a machine learning model, and training the model to obtain a prediction performance index corresponding to the current feature combination; combining the features corresponding to the maximum prediction performance index value into an optimal feature combination, and taking the trained machine learning model corresponding to the maximum prediction performance index value as an optimal machine learning model; and inputting a to-be-predicted feature combination into the trained optimal machine learning model to obtain a gait state prediction result.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of gait recognition technology, in particular to electromyography (EMG) based on machine learning. , EMG) gait recognition and evaluation method and system. Background Art

[0002] With the aging of the global population and the increasing risks of stroke, muscle weakness, falls, etc., rehabilitation and movement assistance for movement disorders have become the focus of social attention. Especially in the field of limb rehabilitation, restoring or improving walking ability is considered to be the core goal of patient rehabilitation. In addition to traditional treatment methods, in recent years, the application of medical robotic devices in the field of movement assistance and rehabilitation has received increasing attention. For example:

[0003] CN105411593B is a method and device for identifying gait tasks, which detects a user's gait pattern based on sensing data, generates gait features of the gait pattern based on the similarity between the gait pattern and each similar gait data extracted from multiple databases, and estimates the gait task corresponding to the gait pattern by applying the gait features to a desired learning model; however, it cannot guarantee the real-time and accuracy of gait recognition.

[0004] Therefore, the demand for the speed and accuracy of gait recognition has been greatly improved, while traditional gait recognition methods face the following problems:

[0005] 1. Poor universality. Manual observation is one of the earliest methods used for gait assessment. Researchers visually observe and analyze the gait of subjects to obtain information about gait characteristics. However, this method is limited by subjectivity, subjective judgment differences, and reliance on experience, and it is difficult to provide objective and accurate gait analysis. In addition, the physiological state of an individual, such as muscle strength, bone density, etc., also has a significant impact on gait. However, existing systems are unable to comprehensively evaluate these physiological factors and cannot effectively distinguish the gaits of people of different ages and physical conditions.

[0006] 2. Sample selection is often not accurate enough and lacks sufficient representativeness. Currently, many systems rely on traditional feature selection strategies, but these methods often fail to deeply consider the unique role and contribution of each feature in gait recognition. Therefore, the detection ability of the system may be limited and it is impossible to fully evaluate the actual effectiveness of different features.

[0007] 3. Insufficient utilization of sample features. In gait detection systems, the utilization of sample features is particularly critical because they are the basis for the system to recognize and understand different gait patterns. However, unfortunately, although the selection of sample features is crucial for gait detection, few systems in reality can make full use of this valuable information. Summary of the invention

[0008] In order to solve the shortcomings of the existing technology, the present invention provides an electromyography gait recognition and evaluation method and system based on machine learning; using biological signals, especially electromyography (EMG), as a key indicator that directly reflects the wearer's muscle activity, the system provides a direct connection with the wearer's movement intention. And a series of methods related to machine learning have confirmed the feasibility of wearable EMG as a gait recognition detection method.

[0009] On the one hand, a machine learning-based electromyography gait recognition and evaluation method is provided, including:

[0010] Acquire human electromyographic signals; construct a training set based on the human electromyographic signals, wherein the training set includes: human electromyographic signals with various known gait labels;

[0011] Extract features from the human electromyographic signals of the training set to obtain several gait features; select features from all gait features to obtain several most representative features; and construct several different feature combinations based on the most representative features;

[0012] Input each feature combination into the machine learning model, train the model, and obtain the prediction performance index corresponding to the current feature combination; take the feature combination corresponding to the maximum prediction performance index value as the optimal feature combination, and take the trained machine learning model corresponding to the maximum prediction performance index value as the optimal machine learning model;

[0013] Acquire the electromyographic signal of the object to be predicted, perform feature extraction on the electromyographic signal of the object to be predicted, obtain the gait characteristics of the object to be predicted, screen out the features consistent with the optimal feature combination from the gait characteristics of the object to be predicted, and thus construct the feature combination to be predicted; input the feature combination to be predicted into the trained optimal machine learning model to obtain the gait state prediction result.

[0014] On the other hand, an electromyography gait recognition and evaluation system based on machine learning is provided, including:

[0015] The acquisition module is configured to: acquire human electromyographic signals; construct a training set based on the human electromyographic signals, wherein the training set includes: human electromyographic signals with various known gait labels;

[0016] The feature extraction module is configured to: extract features from the human electromyographic signals of the training set to obtain a number of gait features; perform feature screening from all gait features to obtain a number of most representative features; and construct a number of different feature combinations based on the most representative features;

[0017] The model training module is configured to: input each feature combination into the machine learning model, train the model, and obtain the prediction performance index corresponding to the current feature combination; take the feature combination corresponding to the maximum prediction performance index value as the optimal feature combination, and take the trained machine learning model corresponding to the maximum prediction performance index value as the optimal machine learning model;

[0018] The gait prediction module is configured to: obtain the electromyographic signal of the object to be predicted, perform feature extraction on the electromyographic signal of the object to be predicted, obtain the gait characteristics of the object to be predicted, screen out the features consistent with the optimal feature combination from the gait characteristics of the object to be predicted, and thus construct the feature combination to be predicted; input the feature combination to be predicted into the trained optimal machine learning model to obtain the gait state prediction result.

[0019] On the other hand, there is also provided an electronic device, comprising:

[0020] a memory for non-transitory storage of computer-readable instructions; and

[0021] a processor for executing the computer readable instructions,

[0022] When the computer-readable instructions are executed by the processor, the method described in the first aspect is executed.

[0023] On the other hand, a storage medium is provided, which non-temporarily stores computer-readable instructions, wherein when the non-temporary computer-readable instructions are executed by a computer, the instructions of the method described in the first aspect are executed.

[0024] On the other hand, a computer program product is provided, comprising a computer program, wherein the computer program is used to implement the method described in the first aspect when running on one or more processors.

[0025] The above technical solution has the following advantages or beneficial effects:

[0026] In order to efficiently process and extract these feature values, the present invention adopts the sliding window method to segment the sample points, and designs a machine learning model for gait recognition based on the time domain features including root mean square (RMS), standard deviation (SD), mean absolute value (MAV), integrated electromyogram (IEMG), zero crossing rate, slope sign change and over-threshold.

[0027] This paper introduces an electromyography gait recognition and evaluation method based on machine learning, which focuses on analyzing the time series data generated by pedestrians during walking. The entire process covers data collection, fine extraction and preprocessing, and training and optimization of gait detection models based on machine learning models. In order to improve the system's real-time responsiveness and adaptability to different users, this paper uses EMG technology to capture muscle activity signals and introduces a sliding window mechanism to dynamically segment signal samples.

[0028] In terms of feature extraction, the present invention adopts a variety of time domain features including root mean square (RMS), standard deviation (SD), mean absolute value (MAV), integrated electromyography (IEMG), zero crossing rate, slope sign change and threshold crossing, aiming to comprehensively and accurately reflect the dynamic characteristics of gait changes. This method not only ensures the real-time performance of gait detection, but also significantly enhances the adaptability of the system to different walking habits and speeds. Through this series of optimization measures, the present invention successfully developed an efficient, real-time and adaptable gait detection solution. The system can accurately identify the current gait stage of the pedestrian and maintain excellent performance at different walking speeds and habits. This innovative technology has brought significant progress in the field of gait assessment and provided strong technical support for future research and applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0030] Figure 1 is a flowchart of an embodiment of the present disclosure;

[0031] Figure 2 is a structural diagram of an artificial neural network according to an embodiment of the present disclosure;

[0032] Figure 3 A system connection diagram of an embodiment of the present disclosure. DETAILED DESCRIPTION

[0033] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0034] Terminology explanation:

[0035] Electromyography (EMG) is used to measure and analyze the electromyographic signals emitted when muscles contract. It represents the activity level of muscles and can be used to study muscle function. By measuring the data indicators of electromyographic response, such as RMS root mean square, iEMG integrated electromyography, MF median frequency, etc., the individual's nerve and muscle function status can be further evaluated.

[0036] Embodiment 1

[0037] This embodiment provides an electromyography gait recognition and evaluation method based on machine learning;

[0038] like Figure 1 As shown, the electromyography gait recognition and evaluation method based on machine learning includes:

[0039] S101: Acquire human electromyographic signals; construct a training set based on the human electromyographic signals, wherein the training set includes: human electromyographic signals with various known gait labels;

[0040] S102: extracting features from the human electromyographic signals of the training set to obtain a number of gait features; screening features from all the gait features to obtain a number of most representative features; and constructing a number of different feature combinations based on the most representative features;

[0041] S103: Input each feature combination into the machine learning model, train the model, and obtain the prediction performance index corresponding to the current feature combination; take the feature combination corresponding to the maximum prediction performance index value as the optimal feature combination, and take the trained machine learning model corresponding to the maximum prediction performance index value as the optimal machine learning model;

[0042] S104: Acquire the electromyographic signal of the object to be predicted, perform feature extraction on the electromyographic signal of the object to be predicted, obtain the gait features of the object to be predicted, screen out features consistent with the optimal feature combination from the gait features of the object to be predicted, and thus construct the feature combination to be predicted; input the feature combination to be predicted into the trained optimal machine learning model to obtain the gait state prediction result.

[0043] Furthermore, if Figure 3 As shown, the step S101: acquiring human electromyographic signals is to acquire human electromyographic signal time series data through electrode patches attached to the surface of human skin.

[0044] Furthermore, the human electromyographic signal refers to the muscle electromyographic signal of the human arm or leg.

[0045] The muscles of the human arm or leg include: tensor fasciae latae, rectus femoris, vastus medialis, semimembranosus, tibialis anterior superior, tibialis anterior inferior, peroneus longus, peroneus brevis and soleus.

[0046] Furthermore, S101: constructing a training set based on human electromyographic signals, the training set comprising: human electromyographic signals with various known gait labels; the gait labels comprising: heel strike category and toe off category; the gait labels comprising: heel strike category, maximum stance flexion category, maximum stance extension category, toe off category and maximum swing flexion category.

[0047] It should be understood that the heel strike category means: Heel strike refers to the moment when the heel first contacts the ground during a walking or running cycle.

[0048] It should be understood that the maximum stance flexion category refers to the degree of flexion of the foot during the stance phase (ie, the phase when the foot is in contact with the ground).

[0049] It should be understood that the maximum stance extension category refers to the state of maximum extension or push-off of the ground achieved by the foot during the stance phase.

[0050] It should be understood that the toe-off category refers to the moment when the toes last leave the ground during the walking or running cycle.

[0051] It should be understood that the maximum swing flexion category refers to the maximum degree of flexion achieved by the foot during the swing process.

[0052] Furthermore, S102: extracting features from the human electromyographic signals of the training set to obtain a plurality of gait features, wherein the plurality of gait features are obtained by sliding the human electromyographic signals with a sliding window, and extracting gait features from the signals in the window each time the window slides.

[0053] Furthermore, the S102: extracting features from the human electromyographic signals of the training set to obtain a plurality of gait features, wherein the plurality of gait features include:

[0054] RMS, standard deviation, mean absolute value, integrated EMG signal, zero crossing rate, slope sign change, and threshold crossing.

[0055] It should be understood that the root mean square refers to the square root of the average value of the squares of the data.

[0056] It should be understood that the standard deviation refers to the square root of the arithmetic mean of the squares of the mean deviations.

[0057] It should be understood that the mean absolute value refers to the average value of the absolute values ​​of each value in the data set.

[0058] It should be understood that the comprehensive electromyographic signal refers to the result of integrating the electromyographic signal within a certain period of time.

[0059] It should be understood that the zero-crossing rate refers to the number of times a signal passes through the zero point in a unit time.

[0060] It should be understood that the slope sign change refers to the number of positive and negative changes in the slope of the signal at consecutive time points.

[0061] It should be understood that the threshold crossing refers to the number of times a signal crosses a set threshold at consecutive time points.

[0062] In the process of data processing and analysis, feature extraction is a crucial step. As an important means to extract meaningful information from electromyographic signals, the accuracy of feature extraction directly affects the accuracy of subsequent gait analysis. In view of the wide application of time domain features (TD) in limb movement recognition, the present invention selects time domain features such as root mean square (RMS), standard deviation (SD), mean absolute value (MAV), integrated electromyographic signal (IEMG), zero crossing rate, slope sign change and threshold crossing to more accurately reflect the dynamic changes of gait.

[0063] In order to effectively process and extract these feature values, the present invention uses a sliding window method to segment the sample points, and the window size is set to 200 samples. This method can perform detailed analysis on different gait stages, thereby improving the accuracy and efficiency of feature extraction.

[0064] Furthermore, the S102: performs feature screening from all gait features to obtain the most representative features, wherein the feature screening uses a principal component analysis algorithm to screen to obtain the most representative features; or, screens according to user selections to screen out features selected by the user, and uses the features selected by the user as the most representative features.

[0065] Furthermore, S102: constructing several different feature combinations based on the most representative features. Assuming that the most representative features are a, b and c, the constructed feature combinations include: (a, b), (a, c), (b, c) and (a, b, c).

[0066] Furthermore, the S103: input each feature combination into the machine learning model, train the model, and obtain the prediction performance index corresponding to the current feature combination, including:

[0067] Each feature combination is used as the input value of the machine learning model, and the gait label corresponding to the current combination is used as the output value of the machine learning model. The machine learning model is trained. When the total loss function value of the machine learning model no longer decreases, the training is stopped to obtain the trained machine learning model.

[0068] Furthermore, the S103: input each feature combination into the machine learning model, train the model, and obtain the prediction performance index corresponding to the current feature combination, including:

[0069] If the gait label includes: a heel strike category and a toe off category, the machine learning model includes: an artificial neural network model, a K nearest neighbor algorithm model or a support vector machine model; the artificial neural network model includes five fully connected layers connected in sequence;

[0070] If the gait labels include: heel strike category, maximum stance flexion category, maximum stance extension category, toe-off category and maximum swing flexion category, then the machine learning model is an artificial neural network model.

[0071] It should be understood that in terms of classification, the present invention performs two classifications based on different gait phases. First, the present invention performs a binary classification experiment, which includes two key gait events, heel strike (HS) and toe off (TO). Subsequently, the present invention expands the classification system into a five-category classification. In the five-category system, the present invention considers five key gait events, including heel strike (HS), maximum stance flexion (MSF), maximum stance extension (MSE), toe off (TO) and maximum swing flexion (MWF), to comprehensively analyze the dynamic changes of gait.

[0072] The artificial neural network model is converted into high-dimensional features of one-dimensional data through fully connected layers to obtain gait feature information. The architecture of this model allows for effective feature extraction, making gait detection more accurate. Gait evaluation model of artificial neural network. The model includes five fully connected layers. The architecture is as follows Figure 2 This network allows extracting local features or features of some muscles from the input gait sequence data, and then synthesizing the local features at a higher level to obtain spatial information.

[0073] The artificial neural network model consists of five fully connected layers to classify and organize the data. In the artificial neural network model, each node (neuron) receives the input signal, sums it after weighting, and applies a nonlinear activation function to generate the output. For the artificial neural network model, the Rectified Linear Unit (ReLU) function is selected as the activation function. The ReLU function performs a nonlinear transformation on the input data, which enhances the responsiveness of the neural network and enables the model to react to the data faster. More importantly, the Batch Normalization function (BatchNorm1d) is used at each level. BatchNorm1d normalizes the input by shifting the mean of each feature dimension to zero and scaling its standard deviation to 1. This stabilization of the input distribution significantly improves the generalization ability of the model and speeds up the training process of the neural network.

[0074] It should be understood that the K nearest neighbor algorithm model means: using the sample-to-sample distance value as a metric, and then selecting the K training samples closest to the test sample based on the calculated sample-to-sample distance, obtaining the category with the largest proportion of training samples from the closest K training samples, and taking the category with the largest proportion of training samples as the category of the test sample.

[0075] Perform preprocessing steps on the dataset, such as using data standardization (StandardScaler) to convert the feature values ​​to a form with a mean of 0 and a standard deviation of 1, which helps many machine learning algorithms, especially those that are sensitive to feature scaling. Use grid search to optimize the hyperparameters of the KNN (K-Nearest Neighbors) algorithm. When K = 50 and use Manhattan distance to calculate their distance, it can more accurately reflect the similarities and differences between samples, thereby obtaining better classification results.

[0076] It should be understood that the support vector machine model refers to finding an optimal decision boundary to maximize the classification margin. The data set is preprocessed, such as using StandardScaler to convert the eigenvalues ​​to a form with a mean of 0 and a standard deviation of 1, which helps many machine learning algorithms, especially those that are sensitive to feature scaling. Grid Search is used to optimize the hyperparameters of the SVM algorithm. When C is set to 20, the kernel function selects a polynomial kernel (poly), and when the kernel function degree is 3, the model obtains better performance.

[0077] The present invention obtains through experiments that for two-category classification, the accuracy of the three models is similar, but for five-category classification, the artificial neural network is the best choice.

[0078] Further, the S103: takes the feature combination corresponding to the maximum prediction performance index value as the optimal feature combination, and takes the trained machine learning model corresponding to the maximum prediction performance index value as the optimal machine learning model; assuming that the prediction accuracy of the first feature combination is q%, the prediction accuracy of the second feature combination is w%, and the prediction accuracy of the third feature combination is p%, and q>w>p, then the first feature combination is the optimal combination, and the machine learning model trained using the first feature combination is the optimal machine learning model.

[0079] Further, the S104: obtains the electromyographic signal of the object to be predicted, performs feature extraction on the electromyographic signal of the object to be predicted, and obtains the gait characteristics of the object to be predicted, and the gait characteristics of the object to be predicted include: root mean square, standard deviation, mean absolute value, comprehensive electromyographic signal, zero crossing rate, slope sign change, and threshold crossing.

[0080] Furthermore, the S104: selects features consistent with the optimal feature combination from the gait features of the object to be predicted, thereby constructing the feature combination to be predicted; if the optimal feature combination is feature a and feature b, then the feature combination to be predicted is also feature a and feature b.

[0081] The present invention provides a wearable electromyography gait recognition evaluation method and system based on machine learning, which relates to the field of gait recognition technology, including: obtaining time series data of people walking; introducing a sliding window method, performing data processing on different muscles and data and generating various feature variables; inputting the processed gait time series data into an artificial neural network, first extracting local features, synthesizing the local features to obtain a spatial feature map, compressing multiple units in the spatial feature map into a single target vector in the fully connected layer, and also using the KNN method to obtain adjacent samples, or using SVM to find a suitable plane. Thus, a suitable training model is obtained; obtaining gait data to be tested, extracting features from the gait data to be tested, and obtaining a number of gait features to be predicted; performing feature selection from a number of gait features to be predicted, and obtaining a combination of features to be predicted; inputting the combination of features to be predicted into the trained machine learning model to obtain a gait recognition result, and proving that artificial neural networks are the best choice for multi-classification problems.

[0082] Embodiment 2

[0083] This embodiment provides an electromyography gait recognition and evaluation system based on machine learning;

[0084] The electromyography gait recognition and evaluation system based on machine learning includes:

[0085] The acquisition module is configured to: acquire human electromyographic signals; construct a training set based on the human electromyographic signals, wherein the training set includes: human electromyographic signals with various known gait labels;

[0086] The feature extraction module is configured to: extract features from the human electromyographic signals of the training set to obtain a number of gait features; perform feature screening from all gait features to obtain a number of most representative features; and construct a number of different feature combinations based on the most representative features;

[0087] The model training module is configured to: input each feature combination into the machine learning model, train the model, and obtain the prediction performance index corresponding to the current feature combination; take the feature combination corresponding to the maximum prediction performance index value as the optimal feature combination, and take the trained machine learning model corresponding to the maximum prediction performance index value as the optimal machine learning model;

[0088] The gait prediction module is configured to: obtain the electromyographic signal of the object to be predicted, perform feature extraction on the electromyographic signal of the object to be predicted, obtain the gait characteristics of the object to be predicted, screen out the features consistent with the optimal feature combination from the gait characteristics of the object to be predicted, and thus construct the feature combination to be predicted; input the feature combination to be predicted into the trained optimal machine learning model to obtain the gait state prediction result.

[0089] It should be noted that the acquisition module, feature extraction module, model training module and gait prediction module correspond to steps S101 to S104 in Embodiment 1, and the examples and application scenarios implemented by the modules are the same as those of the corresponding steps, but are not limited to the contents disclosed in Embodiment 1. It should be noted that the modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0090] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0091] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0092] Embodiment 3

[0093] This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory so that the electronic device executes the method described in the above embodiment one.

[0094] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0095] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0096] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in a processor or an instruction in the form of software.

[0097] The method in the first embodiment can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0098] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0099] Embodiment 4

[0100] This embodiment further provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the method described in the first embodiment is completed.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An electromyography gait recognition and evaluation method based on machine learning, characterized by: include: Obtain human electromyographic signals; Based on human electromyographic signals, a training set is constructed, wherein the training set includes: human electromyographic signals with various known gait labels; Extract features from the human electromyographic signals of the training set to obtain several gait features; select features from all gait features to obtain several most representative features; and construct several different feature combinations based on the most representative features; Input each feature combination into the machine learning model, train the model, and obtain the prediction performance index corresponding to the current feature combination; take the feature combination corresponding to the maximum prediction performance index value as the optimal feature combination, and take the trained machine learning model corresponding to the maximum prediction performance index value as the optimal machine learning model; Acquire the electromyographic signal of the object to be predicted, perform feature extraction on the electromyographic signal of the object to be predicted, obtain the gait characteristics of the object to be predicted, screen out the features consistent with the optimal feature combination from the gait characteristics of the object to be predicted, and thus construct the feature combination to be predicted; input the feature combination to be predicted into the trained optimal machine learning model to obtain the gait state prediction result.

2. The method for electromyography gait recognition and evaluation based on machine learning as claimed in claim 1, characterized in that: Based on human electromyographic signals, a training set is constructed, wherein the training set includes: human electromyographic signals with various known gait labels; the gait labels include: heel strike category and toe drop category; the gait labels include: heel strike category, maximum stance flexion category, maximum stance extension category, toe drop category and maximum swing flexion category.

3. The electromyography gait recognition and evaluation method based on machine learning as claimed in claim 1, characterized in that: The human electromyographic signals of the training set are subjected to feature extraction to obtain a number of gait features, wherein the plurality of gait features are obtained by sliding the human electromyographic signals using a sliding window, and each time a sliding window is used, a gait feature is extracted from the signal in the window.

4. The method for electromyography gait recognition and evaluation based on machine learning as claimed in claim 1, characterized in that: The human electromyographic signals of the training set are subjected to feature extraction to obtain a number of gait features, wherein the gait features include: root mean square, standard deviation, mean absolute value, integrated electromyographic signal, zero crossing rate, slope sign change, and threshold crossing.

5. The method for electromyography gait recognition and evaluation based on machine learning as claimed in claim 1, characterized in that: Each feature combination is input into the machine learning model, and the model is trained to obtain the prediction performance indicators corresponding to the current feature combination, including: Each feature combination is used as the input value of the machine learning model, and the gait label corresponding to the current combination is used as the output value of the machine learning model. The machine learning model is trained. When the total loss function value of the machine learning model no longer decreases, the training is stopped to obtain the trained machine learning model.

6. The method for electromyography gait recognition and evaluation based on machine learning as claimed in claim 1, characterized in that: Each feature combination is input into the machine learning model, and the model is trained to obtain the prediction performance indicators corresponding to the current feature combination, including: If the gait label includes: a heel strike category and a toe off category, the machine learning model includes: an artificial neural network model, a K nearest neighbor algorithm model or a support vector machine model; the artificial neural network model includes five fully connected layers connected in sequence; If the gait labels include: heel strike category, maximum stance flexion category, maximum stance extension category, toe-off category and maximum swing flexion category, then the machine learning model is an artificial neural network model.

7. The method for electromyography gait recognition and evaluation based on machine learning as claimed in claim 6, characterized in that: The K nearest neighbor algorithm model refers to: using the sample-to-sample distance value as a metric, then selecting K training samples closest to the test sample based on the calculated sample-to-sample distance, and obtaining the category with the largest proportion of training samples from the closest K training samples based on the principle of minority obeys majority, and taking the category with the largest proportion of training samples as the category of the test sample.

8. The electromyography gait recognition and evaluation system based on machine learning is characterized by: include: An acquisition module is configured to: acquire a human electromyographic signal; Based on human electromyographic signals, a training set is constructed, wherein the training set includes: human electromyographic signals with various known gait labels; The feature extraction module is configured to: extract features from the human electromyographic signals of the training set to obtain a number of gait features; perform feature screening from all gait features to obtain a number of most representative features; and construct a number of different feature combinations based on the most representative features; The model training module is configured to: input each feature combination into the machine learning model, train the model, and obtain the prediction performance index corresponding to the current feature combination; take the feature combination corresponding to the maximum prediction performance index value as the optimal feature combination, and take the trained machine learning model corresponding to the maximum prediction performance index value as the optimal machine learning model; The gait prediction module is configured to: obtain the electromyographic signal of the object to be predicted, perform feature extraction on the electromyographic signal of the object to be predicted, obtain the gait characteristics of the object to be predicted, screen out the features consistent with the optimal feature combination from the gait characteristics of the object to be predicted, and thus construct the feature combination to be predicted; input the feature combination to be predicted into the trained optimal machine learning model to obtain the gait state prediction result.

9. An electronic device, comprising: a memory for non-transitory storage of computer readable instructions; as well as a processor for executing the computer readable instructions, When the computer-readable instructions are executed by the processor, the method described in any one of claims 1 to 7 is executed.

10. A storage medium, characterized in that: The computer-readable instructions are non-transitory stored, wherein when the non-transitory computer-readable instructions are executed by a computer, the instructions of the method according to any one of claims 1 to 7 are executed.

Citation Information

Cited By

  • Method, system and equipment for predicting photosynthetic rate of protected grapes and storage medium

    CN119760668A

  • A method, system, device and storage medium for predicting the photosynthetic rate of greenhouse grapes

    CN119760668B

  • Snore detection method and system based on intelligent bed

    CN119867667A

  • A Snoring Detection Method and System Based on an Intelligent Bed

    CN119867667B

  • Skin penetrating peptide classifier construction method and device based on machine learning, equipment and medium

    CN121658989A