Lower limb movement gait recognition method and system based on feature fusion

Abstract

The invention discloses a lower limb movement gait recognition method and system based on feature fusion, and belongs to the technical field of gait recognition. Comprising the following steps: acquiring a surface electromyogram signal and a motion inertia signal of a to-be-predicted object; carrying out feature extraction on the surface electromyogram signal and the motion inertia signal by utilizing wavelet transform, obtaining gait features and screening an optimal feature combination; and processing the optimal feature combination through the trained gait recognition model to obtain a gait recognition result. Complex multi-modal data can be effectively processed and analyzed, so that high-precision gait recognition and evaluation are realized; the problem that the accuracy of gait evaluation needs to be improved due to the fact that single signal processing is difficult to comprehensively reflect a complex gait dynamic process in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of gait recognition, and in particular to a method and system for lower limb movement gait recognition based on feature fusion. Background Art

[0002] The statements in this section merely mention background art related to the present invention and do not necessarily constitute prior art.

[0003] Wearable devices are increasingly being used in the field of lower limb rehabilitation and motion assessment. These devices can provide important feedback information for rehabilitation therapy by monitoring and analyzing the user's gait data in real time. However, traditional gait recognition technology mainly relies on a single type of sensor, such as force sensors or inertial measurement units (IMUs), which limits the precision and accuracy of gait recognition. Force sensors mainly capture ground reaction forces, while IMUs capture linear and angular velocity information of motion through accelerometers and gyroscopes. Although these sensors can provide valuable gait information, a single sensor is often difficult to fully and accurately reflect the complex gait dynamic process.

[0004] In recent years, surface electromyography (sEMG) has received extensive attention as a signal that directly reflects muscle electrical activity. sEMG signals can capture muscle activation patterns, directly reflect the wearer's movement intentions, and are closely related to muscle activity during gait. However, due to the characteristics of sEMG signals such as susceptibility to noise interference and poor stability, there are certain problems in relying solely on sEMG signals for gait recognition, which affects the accuracy of gait recognition.

[0005] However, due to the different data types of surface EMG and IMU data, the sampling frequency and data format may be inconsistent, which makes the combination of surface EMG and IMU data in lower limb gait recognition face challenges in signal synchronization and data fusion.

[0006] In addition, sEMG signals are susceptible to noise interference, such as electromagnetic interference, muscle fatigue, etc. These noises may mask the true muscle activity signals and thus affect the accuracy of gait recognition. Summary of the invention

[0007] In order to address the deficiencies of the prior art, the present invention provides a method, system, electronic device, computer-readable storage medium and computer program product for lower limb motion gait recognition based on feature fusion, which fuses surface electromyography signals and motion inertial signals to provide comprehensive and accurate information for gait recognition and achieve high-precision gait recognition and evaluation.

[0008] In a first aspect, the present invention provides a method for lower limb movement gait recognition based on feature fusion;

[0009] A method for lower limb motion gait recognition based on feature fusion, comprising:

[0010] Acquire surface electromyography signals and motion inertia signals of the object to be predicted;

[0011] Using wavelet transform to extract features from the surface electromyography signal and the motion inertia signal, obtain gait features and select the optimal feature combination;

[0012] The optimal feature combination is processed through the trained gait recognition model to obtain the gait recognition result.

[0013] In some embodiments, before extracting features from the surface electromyography signal and the motion inertia signal using wavelet transform, the method further includes:

[0014] The surface electromyography signal and the motion inertia signal are filtered, the start and end points of the gait cycle are determined according to the acceleration peak or angular velocity change in the motion inertia signal, and the continuous surface electromyography signal and the motion inertia signal are divided into single step cycles.

[0015] In some embodiments, the use of wavelet transform to extract features from the surface electromyography signal and the motion inertia signal is specifically as follows: performing wavelet transform on the surface electromyography signal to extract wavelet coefficients, energy features, and entropy features; performing wavelet transform on the motion inertia signal to extract time-frequency domain features of acceleration signals and angular velocity signals.

[0016] In some embodiments, training the gait recognition model specifically includes: constructing a training set using surface electromyography signals and motion inertia signals annotated with multiple gait type labels, and optimizing the gait recognition model using the training set through a cross entropy loss function and an Adam optimizer.

[0017] In some embodiments, feature screening of a plurality of gait features specifically includes:

[0018] Based on statistical methods, several gait features are preliminarily screened, and the correlation between each gait feature and the gait category is evaluated. Feature screening is performed based on the correlation to obtain the first gait feature data set;

[0019] The first gait feature data set is screened by a model-based recursive feature elimination method to determine a second gait feature data set;

[0020] The second gait feature dataset is sparsely processed through L1 regularization to obtain the optimal feature combination.

[0021] In some embodiments, the gait recognition model is a multi-layer perceptron model.

[0022] In a second aspect, the present invention provides a lower limb movement gait recognition system based on feature fusion;

[0023] A lower limb motion gait recognition system based on feature fusion, comprising:

[0024] The wearable device body is used to collect the surface electromyography signal and motion inertia signal of the object to be predicted;

[0025] The data processing module is used to obtain the surface electromyography signal and the motion inertia signal of the object to be predicted; use wavelet transform to extract the features of the surface electromyography signal and the motion inertia signal, obtain gait features and screen the optimal feature combination; and process the optimal feature combination through the trained gait recognition model to obtain the gait recognition result.

[0026] In a third aspect, the present invention provides an electronic device;

[0027] An electronic device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-mentioned method for lower limb motion gait recognition based on feature fusion.

[0028] In a fourth aspect, the present invention provides a computer-readable storage medium;

[0029] A computer-readable storage medium stores a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned method for lower limb motion gait recognition based on feature fusion.

[0030] In a fifth aspect, the present invention provides a computer program product;

[0031] A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned method for lower limb movement gait recognition based on feature fusion.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. The technical solution provided by the present invention combines surface electromyography signals and motion inertia signals to provide more comprehensive and accurate gait information for gait recognition; combined with machine learning technology, it can effectively process and analyze complex multimodal data, thereby achieving high-precision gait recognition and evaluation.

[0034] 2. The technical solution provided by the present invention fully integrates sEMG features and IMU features to utilize the complementary information of the two signal sources for gait recognition of lower limb movements. It has a higher feature dimension, can more comprehensively describe the spatiotemporal characteristics of gait, and provide richer feature information for gait recognition, thereby improving the accuracy of gait recognition.

[0035] 3. The technical solution provided by the present invention integrates the data collection, processing and gait recognition of electromyography and inertial measurement units into an integrated wearable device, which can provide high-precision, real-time gait evaluation, timely monitoring and feedback of gait information; the integrated accelerometer, gyroscope and sEMG module can simultaneously capture motion and bioelectric signals, thereby providing comprehensive data collection. The device has a built-in 32MB non-volatile memory, which can store data locally and transmit it wirelessly via Bluetooth, greatly improving the flexibility of data management.

[0036] 4. The technical solution provided by the present invention is lightweight and easy to wear, making gait data collection and real-time analysis more convenient, and improving data accuracy and measurement reliability. This wearable gait recognition and evaluation device for lower limbs that integrates sEMG and IMU based on machine learning has broad application prospects. It can be used by rehabilitation centers, medical institutions and individual users to monitor and improve gait abnormalities, prevent and treat gait-related diseases, and improve individual athletic ability and quality of life. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] 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.

[0038] Figure 1 A schematic diagram of a flow chart of a method for lower limb movement gait recognition based on feature fusion provided in an embodiment of the present invention;

[0039] Figure 2 A schematic diagram of the structure of a lower limb movement gait recognition system based on feature fusion provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] 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 in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0041] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0042] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0043] Embodiment 1

[0044] Existing gait assessment only uses single sensor information for processing, which cannot fully and accurately reflect the complex gait dynamic process; therefore, the present invention provides a lower limb motion gait recognition method based on feature fusion, which organically combines surface electromyography signals and motion inertia signals, extracts gait features to reflect comprehensive and accurate gait information.

[0045] Next, combine Figure 1 , a method for recognizing lower limb movement gait based on feature fusion disclosed in this embodiment is described in detail. The method for recognizing lower limb movement gait based on feature fusion is applied to a processing terminal and includes the following steps:

[0046] S1. Acquire the surface electromyography signal and motion inertia signal of the object to be predicted, where the motion inertia signal includes an acceleration signal and an angular acceleration signal.

[0047] In this embodiment, the object to be predicted is made to wear a wearable device integrated with a sEMG sensor and an IMU sensor to collect surface electromyographic signals and motion inertial signals. The sEMG sensor is used to collect surface electromyographic signals, and the IMU sensor includes a three-axis accelerometer and a six-axis gyroscope, which are respectively used to measure acceleration and angular acceleration in three-dimensional space. The data transmission unit collects raw data from the sEMG and IMU sensors, and transmits the collected data to the processing end via Bluetooth.

[0048] S2. Use wavelet transform to extract features of surface electromyography signals and motion inertia signals, obtain gait features and screen the optimal feature combination.

[0049] Since surface electromyography information has the characteristics of being susceptible to noise interference and having poor stability, as an implementation method, before executing S2, it also includes: filtering, denoising, and normalizing the surface electromyography signal and the motion inertia signal, determining the start and end points of the gait cycle according to the acceleration peak or angular velocity change in the motion inertia signal, and dividing the continuous surface electromyography signal and motion inertia signal into single-step cycles.

[0050] The start and end points of the gait cycle mark the occurrence of key gait events, such as heel strike and toe off. By determining these points, the various phases of gait (swing phase, support phase, etc.) can be analyzed more accurately, and the features within each gait cycle can be extracted.

[0051] As an implementation mode, S2 specifically includes:

[0052] S201, performing wavelet transform on the surface electromyography signal and the motion inertia signal to extract time-frequency domain features.

[0053] Specifically, the surface electromyographic signal is subjected to wavelet transform to extract wavelet coefficients, energy characteristics, entropy characteristics, etc. The intensity, frequency and time changes of muscle activation can be reflected through the above time-frequency domain characteristics.

[0054] Wavelet transform is performed on the acceleration signal and angular acceleration signal to extract the time-frequency domain characteristics of the acceleration signal and angular acceleration signal, such as wavelet energy, wavelet entropy, wavelet coefficients at a specific scale, etc., which are used to capture the dynamic changes of gait.

[0055] S202 , dividing the motion inertia signal into gait phases in combination with the gait events, and extracting the accelerometer features and angular velocity meter features of the motion inertia signal.

[0056] Specifically, the motion inertial signal is divided into gait phases in combination with gait events (such as the maximum plantar pressure point), and time domain features such as gait cycle, step frequency, stride, and swing time are extracted.

[0057] Based on this, the gait features obtained by extracting features from the surface electromyography signal are the gait features corresponding to the surface electromyography signal, and the gait features obtained by extracting features from the motion inertia signal are the gait features corresponding to the motion inertia signal.

[0058] In order to ensure the accuracy of feature extraction and reduce noise interference, before executing S201, the sEMG signal and the IMU signal need to be preprocessed, including:

[0059] Step 1: Apply a bandpass filter to filter the sEMG signal and limit the bandwidth of the sEMG signal to 20-450 Hz to effectively remove low-frequency drift and high-frequency noise.

[0060] Step 2: In order to eliminate the differences in muscle activity intensity between different individuals, the maximum voluntary contraction method (MVC) is used to normalize the filtered sEMG signal and convert it into a relative value. This normalization process can more accurately compare the electromyographic activity of different individuals.

[0061] Step 3: Filter the IMU signal through a low-pass filter to remove high-frequency noise to ensure the smoothness of the signal.

[0062] In this embodiment, the cut-off frequency is selected to be below 20 Hz to retain the main components of the gait motion.

[0063] Step 4: Perform gravity compensation on the filtered IMU signal based on the complementary filtering method of posture estimation to ensure that the IMU signal can accurately reflect the dynamic characteristics of the gait.

[0064] In this step, considering that the acceleration signal recorded by the IMU sensor contains a gravity component, the IMU signal is gravity compensated to extract the pure motion acceleration.

[0065] Step 5: Use the dynamic time warping (DTW) method to synchronize the sEMG and IMU signals to ensure better alignment in the time dimension.

[0066] S203: Screening a number of gait features to obtain an optimal feature combination.

[0067] Considering that there are a large number of gait features extracted by wavelet transform, and that there are redundant features in the recognition of lower limb movement gait, it is necessary to perform feature screening. As an implementation method, S203 specifically includes:

[0068] S2031. Preliminarily screen several gait features based on statistical methods, evaluate the correlation between each gait feature and the gait category, and screen the features based on the correlation to obtain the first gait feature data set.

[0069] Specifically, first, remove features with low variance, which vary little between samples and have limited contribution to classification.

[0070] Then, the correlation between each feature and the gait category was evaluated by methods such as the chi-square test, F test, or mutual information, and the features with higher scores were selected as the first gait feature dataset for the next step of screening.

[0071] S2032. Screen the first gait feature data set by using a model-based recursive feature elimination method to determine a second gait feature data set.

[0072] Specifically, the first step feature dataset is recursively screened using the RFE method based on linear support vector machine (SVM) or random forest. After each model training, the features with lower weights are removed and the model is retrained until the remaining features reach the specified number. In the recursive process, the model performance after each feature removal is recorded, and the final retained features are determined based on the feature importance score to form the second gait feature dataset.

[0073] Next, taking SVM as an example, S2032 is described in detail. The specific process is as follows:

[0074] (1) The SVM using the linear kernel function is trained using the first step feature data set to obtain the weight vector ω.

[0075] (2) According to the absolute value of the weight vector corresponding to each gait feature, determine the importance score corresponding to each gait feature; here, the larger the absolute value of the weight vector, the greater the contribution of the gait feature to the recognition of lower limb movement gait. The importance score is expressed as:

[0076] Importance(f i )=|ω i |;

[0077] In the formula, f i represents the i-th gait feature.

[0078] (3) Sort the gait features from low to high according to their importance scores and remove the gait features with the lowest scores.

[0079] In some embodiments, random forests can also be used to recursively screen the first step feature data set. The specific process is as follows:

[0080] (1) Use the first step feature dataset to train the random forest model.

[0081] (2) Based on the mean decrease in impurity, the sum of the impurity reduction caused by each gait feature in all decision trees is calculated to determine the importance score corresponding to each gait feature. It can be expressed as:

[0082]

[0083] Where N trees represents the number of trees in the random forest model, represents the set of nodes in the t-th tree that are split using the gait feature fi, N n Represents the number of samples passing through node n, N total represents the total number of samples, and ΔI(n) represents the reduction in impurity at node n.

[0084] (3) Sort the gait features from low to high according to their importance scores and remove the gait features with the lowest scores.

[0085] Based on this, in the recursive process, the model performance after each feature removal is recorded. By analyzing the relationship between model performance and the number of features, the feature combination with the best performance is selected to form the second gait feature dataset. This process helps to improve the generalization ability of the model, reduce the risk of overfitting, and reduce computational complexity.

[0086] S2033. Perform sparse processing on the second gait feature data set through L1 regularization.

[0087] Specifically, the second gait feature data set is used to train the LASSO regression model. In the LASSO regression model, the model assigns a regression coefficient to each gait feature in the second gait feature data set, and these regression coefficients reflect the contribution of each gait feature to the model prediction.

[0088] By introducing the L1 regularization term, the model will force the regression coefficients of some features to approach zero. Gait features with regression coefficients approaching zero are considered to contribute less to the model and can be removed. Finally, important gait features with larger regression coefficients are retained to obtain the third gait feature dataset.

[0089] S2034. Concatenate and fuse the gait features corresponding to the sEMG signal and the gait features corresponding to the IMU signal in the third gait feature data set to obtain a fused feature vector to form an optimal feature combination.

[0090] For example, if the gait feature vector corresponding to the sEMG signal is represented by F sEMG =[f 1 ,f 2 ......,f m ], the gait feature vector corresponding to the IMU signal is expressed as F IMU =[f m+1 ,f m+2 ,...,f n ], then the fused feature vector is expressed as:

[0091] F 融合 =[f 1 ,f 2 ,...,f m ,f m+1 ,f m+2 ,...,f n ].

[0092] In this step, the optimal feature combination finally obtained contains information from two signal sources, has a higher feature dimension, and can more comprehensively describe the spatiotemporal characteristics of gait.

[0093] S3. The optimal feature combination is processed through a trained gait recognition model to obtain a gait recognition result; the gait recognition model is a trained multi-layer perceptron (MLP) model, the input is the optimal feature combination, and the output is the gait recognition result, such as different stages and sub-stages of gait.

[0094] In this embodiment, a multi-layer perceptron (MLP) model includes an input layer, two hidden layers (including 64 and 32 neurons respectively) and an output layer for processing gait classification tasks. During the training process, surface electromyographic signals and motion inertial signals labeled with various gait type labels are used and processed as described in S2 to construct a training set. The model is optimized using the training set through a cross entropy loss function and an Adam optimizer to ensure that the model can accurately identify gait patterns of different categories.

[0095] Embodiment 2

[0096] Combination Figure 2 The present embodiment discloses a lower limb motion gait recognition system based on feature fusion, including a wearable device body and a data processing module. A surface electromyography (sEMG) sensor, an inertial measurement unit (IMU) and a feedback unit are installed on the wearable device body; the sEMG sensor is used to collect surface electromyography signals and record the electrical activities of the leg muscles in the gait cycle, such as the activities of the quadriceps and gastrocnemius muscles; the IMU is used to collect motion inertial signals and record the changes in acceleration and angular velocity in the gait cycle, such as key events such as heel landing and toe lift-off. By using sEMG and IMU data in coordination, each stage of the gait cycle can be analyzed in detail to evaluate the stability and coordination of the gait. Similarly, combining sEMG and IMU data can provide real-time feedback for rehabilitation training, help adjust the training plan, and improve the rehabilitation effect.

[0097] In this embodiment, the number of surface electromyography sensors is 8 and the number of inertial measurement units is 4.

[0098] In order to avoid data interference between the sEMG sensor and the IMU, as an implementation method, the wearable device body is designed as a double layer, and the sEMG sensor and the IMU are installed on different layers to improve the acquisition accuracy and avoid mutual interference.

[0099] Furthermore, a wireless transmission unit is installed on the wearable device body, and the collected data is transmitted to the data processing module through the wireless transmission unit to ensure the real-time and stability of data transmission.

[0100] The data processing module is used to obtain the surface electromyography signal and motion inertia signal of the object to be predicted; use wavelet transform to extract features of the surface electromyography signal and motion inertia signal, obtain gait features and screen the optimal feature combination; process the optimal feature combination through the trained gait recognition model to obtain gait recognition results.

[0101] Here, the workflow of the data processing model is the same as that described in Example 1 and will not be repeated here.

[0102] In this embodiment, the wearable device body is made of 3D printed PLA (polylactic acid).

[0103] Embodiment 3

[0104] Embodiment 3 of the present invention provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps of the above-mentioned method for lower limb motion gait recognition based on feature fusion are completed.

[0105] Embodiment 4

[0106] Embodiment 4 of the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above-mentioned method for lower limb motion gait recognition based on feature fusion are completed.

[0107] Embodiment 5

[0108] Embodiment 5 of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned method for lower limb motion gait recognition based on feature fusion.

[0109] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0112] 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.

[0113] 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. A method for lower limb gait recognition based on feature fusion, characterized in that: include: Acquire surface electromyography signals and motion inertia signals of the object to be predicted; Using wavelet transform to extract features from the surface electromyography signal and the motion inertia signal, obtain gait features and select the optimal feature combination; The optimal feature combination is processed through the trained gait recognition model to obtain the gait recognition result.

2. The method for lower limb movement gait recognition based on feature fusion as claimed in claim 1, characterized in that: Before extracting features from the surface electromyographic signal and the motion inertia signal using wavelet transform, the method further includes: The surface electromyography signal and the motion inertia signal are filtered, the start and end points of the gait cycle are determined according to the acceleration peak or angular velocity change in the motion inertia signal, and the continuous surface electromyography signal and the motion inertia signal are divided into single step cycles.

3. The method for lower limb movement gait recognition based on feature fusion as claimed in claim 1, characterized in that: The method of extracting features from the surface electromyographic signal and the motion inertia signal by using wavelet transform specifically comprises: performing wavelet transform on the surface electromyographic signal to extract wavelet coefficients, energy features and entropy features; The motion inertia signal is subjected to wavelet transformation to extract the time-frequency domain features of the acceleration signal and the angular velocity signal.

4. The method for lower limb movement gait recognition based on feature fusion as claimed in claim 1, characterized in that: The specific step of training the gait recognition model is as follows: constructing a training set using gait features annotated with a plurality of gait type labels, and optimizing the gait recognition model using the training set through a cross entropy loss function and an Adam optimizer.

5. The method for lower limb movement gait recognition based on feature fusion as claimed in claim 4, characterized in that: Feature screening of several gait features specifically includes: Based on statistical methods, several gait features are preliminarily screened, and the correlation between each gait feature and the gait category is evaluated. Feature screening is performed based on the correlation to obtain the first gait feature data set; The first gait feature data set is screened by a model-based recursive feature elimination method to determine a second gait feature data set; The second gait feature dataset is sparsely processed through L1 regularization to obtain the optimal feature combination.

6. The method for lower limb movement gait recognition based on feature fusion as claimed in claim 1, characterized in that: The gait recognition model is a multi-layer perceptron model.

7. The lower limb movement gait recognition system based on feature fusion is characterized by: include: The wearable device body is used to collect the surface electromyography signal and motion inertia signal of the object to be predicted; The data processing module is used to obtain the surface electromyography signal and the motion inertia signal of the object to be predicted; use wavelet transform to extract the features of the surface electromyography signal and the motion inertia signal, obtain gait features and screen the optimal feature combination; and process the optimal feature combination through the trained gait recognition model to obtain the gait recognition result.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method for lower limb motion gait recognition based on feature fusion as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the method for lower limb movement gait recognition based on feature fusion described in any one of claims 1-6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method for lower limb movement gait recognition based on feature fusion described in any one of claims 1-6 are implemented.

Images