Gait trajectory prediction method based on dendritic neural network and related device

Abstract

The present application relates to the technical field of biological signal processing, in particular to a gait trajectory prediction method based on a dendritic neural network and related equipment, which comprises the following steps: first, acquiring sEMG signals and EEG signals of muscle groups during double-leg movement, and preprocessing the signals; then inputting the preprocessed signals into a gait trajectory prediction model to predict double-leg gait trajectories and obtain knee joint movement trajectories. The gait trajectory prediction model adopts a dendritic neural network model, which can fuse the input sEMG signals and EEG signals, decode the fused data, and finally obtain double-leg knee joint movement trajectories based on the decoded data, thereby improving the accuracy of gait trajectory prediction.

Description

Technical Field

[0001] This invention relates to the field of biosignal processing technology, specifically to a gait trajectory prediction method and related equipment based on dendritic neural networks. Background Technology

[0002] For normal individuals, the elderly, or exoskeleton wearers with normal movement on one side of their body, decoding the wearer's gait trajectory allows for active control of the lower limb exoskeleton, achieving human-machine collaboration. Based on the mechanism of human gait generation, gait information is contained in signals such as plantar pressure, joint angles, surface electromyography (sEMG), and electroencephalogram (EEG). Currently, the signals used for gait trajectory decoding are mainly kinematic signals such as joint angles. However, these kinematic signals are generated after actual movement occurs and cannot predict gait trajectory, making them difficult to directly use for lower limb exoskeleton control.

[0003] Since electromyography (EMG) signals are generated before the actual gait, they can theoretically be used to predict gait trajectories. However, the amount of gait information that can be decoded from EMG signals is small, which makes it difficult to improve the accuracy of gait trajectory prediction. Therefore, it is urgent to propose a method that can decode gait information from EMG signals to achieve continuous real-time control of the lower limb exoskeleton. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a gait trajectory prediction method and related equipment based on dendritic neural networks to address the shortcomings of the prior art, thereby solving the technical problems of low timeliness and accuracy in lower limb gait trajectory decoding.

[0005] The objective of this invention is achieved through the following technical solutions: In a first aspect, the present invention provides a gait trajectory prediction method based on a dendritic neural network, comprising: Acquire sEMG signals of muscle groups, kinematic signals of major joints of the lower limbs, and EEG signals of the cerebral cortex during leg movements; The sEMG signal, kinematic signal, and cerebral cortex EEG signal are preprocessed; The preprocessed sEMG signal, kinematic signal and EEG signal are input into the gait trajectory prediction model to predict the gait trajectory of both legs and obtain the knee joint movement trajectory of both legs. The gait trajectory prediction model uses a dendritic neural network model. The dendritic neural network model fuses the input sEMG signal and EEG signal, decodes the fused data, and obtains the knee joint movement trajectory of both legs based on the decoded data.

[0006] As a further improvement of the present invention, the preprocessing process includes: After resampling the EEG signal, a denoising operation is performed based on the sampled EEG signal. The denoising operation includes at least re-reference, bandpass filtering, removal of independent components related to electrooculography, and removal of motion artifacts in sequence on the sampled EEG signal. The sEMG signal is subjected to bandpass filtering and wavelet denoising.

[0007] As a further improvement of the present invention, the bandpass filter used for bandpass filtering of the sEMG signal is a two-stage IIR filter based on Butterworth filter design, with a filtering frequency of 1-3Hz delta band.

[0008] As a further improvement of the present invention, it also includes normalizing the sEMG signal and EEG signal obtained after preprocessing.

[0009] As a further improvement of the present invention, the dendritic neural network model is used to simulate the dendritic structure of neurons, and the prediction method of the dendritic neural network model is as follows:

[0010] In the formula, For the first The output of a dendritic branch layer For the first The output of a dendritic branch layer For the first From the first dendritic branch layer to the... The weight matrix of each dendritic branch layer For Hadama accumulation.

[0011] As a further improvement of the present invention, the dendritic neural network model includes a data fusion module, which is used to perform parallel data fusion of the input sEMG signal and EEG signal; the parallel data fusion includes: When sEMG and EEG signals are simultaneously input into a dendritic neural network model, and the label in the dendritic neural network is set to the knee joint motion trajectory, the dendritic neural network can directly learn to decode the knee joint motion trajectory from the sEMG and EEG signals.

[0012] As a further improvement of the present invention, the data fusion module is used to perform concatenated data fusion of the input sEMG signal and EEG signal; the concatenated data fusion includes: Two cascaded dendritic neural networks are set up. The EEG signal is input into the first dendritic neural network, and the label of the first dendritic neural network is set as the sEMG signal. The sEMG signal is obtained by decoding based on the first dendritic neural network. The decoded sEMG signal is used as the input of the second dendritic neural network, and the label of the second dendritic neural network is set as the knee joint motion trajectory. The knee joint motion trajectory is output based on decoding the sEMG signal.

[0013] Secondly, the present invention provides a gait trajectory prediction system based on a dendritic neural network, comprising: The motion data acquisition module acquires sEMG and EEG signals of muscle groups during leg movements. The data preprocessing module preprocesses the sEMG and EEG signals; The gait trajectory prediction module inputs the preprocessed sEMG and EEG signals into the gait trajectory prediction model to predict the gait trajectory of both legs and obtain the knee joint movement trajectory of both legs. The gait trajectory prediction model uses a dendritic neural network model. The dendritic neural network model fuses the input sEMG signal and EEG signal, decodes the fused data, and obtains the knee joint movement trajectory of both legs based on the decoded data.

[0014] Thirdly, the present invention provides a computer-readable storage medium for storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform the above-described gait trajectory prediction method based on a dendritic neural network.

[0015] Fourthly, the present invention provides a computing device, comprising: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including steps for performing the above-described dendritic neural network-based gait trajectory prediction method.

[0016] The beneficial effects of this invention are as follows: This invention provides a gait trajectory prediction method based on dendritic neural networks. By acquiring sEMG (surface electromyography) signals of muscle groups and EEG (electroencephalography) signals of major joints of the lower limbs during leg movement, and after preprocessing, the signals are input into a gait trajectory prediction model using a dendritic neural network model. This dendritic neural network model fuses and decodes the input sEMG and EEG signals, and obtains the knee joint movement trajectory of both legs based on the decoded data. Thus, by utilizing the fusion and decoding capability of the dendritic neural network model for sEMG and EEG signals, the effective prediction of the knee joint movement trajectory of both legs is achieved, solving the technical problem of how to accurately obtain the knee joint movement trajectory of both legs based on sEMG and EEG signals. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the gait trajectory prediction method based on dendritic neural network in an embodiment of the present invention; Figure 2 This is a schematic diagram of the brain-muscle electroconvulsive series-parallel fusion strategy in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a computer device in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives and technical solutions of this invention clearer and easier to understand, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0020] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. The described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] Example 1 This embodiment provides a gait trajectory prediction method based on dendritic neural networks, such as... Figure 1As shown, the main steps of this method are as follows: acquiring sEMG signals of muscle groups, kinematic signals of major joints of the lower limbs, and EEG signals of the cerebral cortex during the movement of the experimental subject's legs; preprocessing the sEMG signals, kinematic signals, and EEG signals of the cerebral cortex; inputting the preprocessed sEMG signals, kinematic signals, and EEG signals into the gait trajectory prediction model to predict the gait trajectory of the legs and obtain the knee joint movement trajectory of the legs.

[0022] In this embodiment, the gait trajectory prediction model uses a dendritic neural network model. The dendritic neural network model fuses the input sEMG signal and EEG signal, decodes the fused data, and obtains the knee joint movement trajectory of both legs based on the decoded data.

[0023] In this embodiment, sEMG signals of muscle groups and EEG signals of major joints of the lower limbs are acquired during leg movement. These signals are preprocessed and then input into a gait trajectory prediction model using a dendritic neural network. The model fuses the input sEMG and EEG signals and decodes the fused data. Finally, based on the decoded data, the knee joint movement trajectory of both legs is accurately predicted. This embodiment achieves the effective integration and utilization of multi-source biological signals, significantly improving the accuracy and reliability of gait trajectory prediction, and providing more precise technical support for related motion analysis and rehabilitation treatment.

[0024] Furthermore, in this embodiment, there are several sEMG and EEG signals.

[0025] Furthermore, in this embodiment, the EEG signal is first resampled, and then denoising operations such as rereference, bandpass filtering, removal of independent components related to electrooculography (EOG) and removal of motion artifacts are sequentially performed on the resampled EEG signal. At the same time, bandpass filtering and wavelet denoising operations are performed on the sEMG signal. Using the processing steps designed for the EEG signal and sEMG signal respectively, the interference components (such as EOG interference, motion artifacts and non-target frequency signals in the EEG signal, and noise in the sEMG signal) present in the two types of signals are targeted and processed. This achieves effective noise reduction and signal quality optimization for the EEG signal and sEMG signal, solves the technical problem that the noise and interference components in the EEG signal and sEMG signal affect the usability of the signal, and provides high-quality signal input support for subsequent data applications based on these two types of signals.

[0026] Furthermore, in this embodiment, the bandpass filter used for bandpass filtering is a two-stage IIR filter designed based on a Butterworth filter, with a filtering frequency of 1-3Hz in the delta band, which has a high content of motion information.

[0027] This embodiment further includes normalizing the preprocessed sEMG and EEG signals. The normalized sEMG and EEG signals are then input into the gait trajectory prediction model.

[0028] The gait trajectory prediction model is a pre-trained model. The construction and training process of this model includes: First, a dendritic neural network model is constructed as the gait trajectory prediction model. The dendritic neural network designed in this embodiment simulates the function of the dendritic structure of human neurons and is a simple white-box model, which can be represented as follows:

[0029] In the formula, For the first The output of a dendritic branch layer For the first The output of a dendritic branch layer For the first From the first dendritic branch layer to the... The weight matrix of each dendritic branch layer For Hadama accumulation.

[0030] The dendritic neural network model mainly consists of an input layer, a dendritic branching layer, a cell body integration layer, and an output layer. The input layer corresponds to the dendritic terminals and receives external sEMG and EEG signals; each input node corresponds to a raw feature. The dendritic branching layer groups the features from the input layer, with each branch processing a set of locally relevant features and performing local integration. The cell body integration layer summarizes the local outputs from all dendritic branching layers, performing global integration to establish a mapping relationship between signal features and knee joint motion trajectories. The output layer decodes the mapping relationship from the cell body integration layer, extracting the core information related to the knee joint motion trajectory and outputting the knee joint motion trajectories of both legs.

[0031] The working principle of the dendritic neural network model is as follows: It receives preprocessed sEMG and EEG signals, and the input layer distributes the signals to different dendritic branches. Each dendritic branch processes the assigned local signal, such as calculating the local temporal features of the sEMG, the specific frequency band energy of the EEG, or the co-variation pattern of the two, filtering effective features through local weighting and activation. The cell body layer summarizes the processing results of all branches, performs global integration and key information extraction, and establishes a mapping relationship between signal features and knee joint motion trajectory. After decoding by the output layer, the integrated information is transformed into specific motion trajectory data of the knee joints of both legs.

[0032] After creating the dendritic neural network model, the sEMG signals of the muscle groups related to the movement of the legs of the experimental subjects, the movement trajectories of the main joints of the lower limbs, and the EEG signals were obtained.

[0033] The acquired EEG, sEMG, and knee joint motion trajectory signals will be resampled. The EEG signals will undergo denoising processes such as rereference, bandpass filtering, removal of independent components related to electrooculography, and removal of motion artifacts. The sEMG signals will undergo preprocessing such as bandpass filtering and wavelet denoising. The preprocessed EEG, sEMG, and knee joint motion trajectory signals will then be normalized. The dendritic neural network model is initialized, including setting the number of samples in a batch, the number of network layers, etc., and then trained based on the collected EEG, sEMG, and knee joint motion trajectory signals.

[0034] Furthermore, this embodiment also includes a data fusion module in the dendritic neural network model. Each fusion module is used to implement two fusion methods: parallel fusion and series fusion.

[0035] The data fusion module performs parallel data fusion on the input sEMG and EEG signals. Specifically, the parallel data fusion operation involves simultaneously inputting the sEMG and EEG signals into a single dendritic neural network model, setting the label in the dendritic neural network to the knee joint motion trajectory. This allows the dendritic neural network to directly learn and decode the knee joint motion trajectory from the input sEMG and EEG signals. By leveraging parallel data fusion, the synchronous input and collaborative utilization of sEMG and EEG signal information are achieved. This allows the dendritic neural network to directly establish a mapping relationship between the two signals and the knee joint motion trajectory, solving the problem of how to efficiently decode the knee joint motion trajectory based on sEMG and EEG signals. It ensures the synchronous functioning of the two signal information during the knee joint motion trajectory decoding process, improving the directness and efficiency of trajectory decoding.

[0036] Furthermore, the data fusion module performs concatenated data fusion on the input sEMG and EEG signals. Specifically, the concatenated data fusion operation involves setting up two concatenated dendritic neural networks. The EEG signal is input into the first dendritic neural network, and its label is set to the sEMG signal. The sEMG signal is then decoded based on this first dendritic neural network. This decoded sEMG signal is then used as the input to the second dendritic neural network, and its label is set to the knee joint motion trajectory. The second dendritic neural network decodes the decoded sEMG signal to output the knee joint motion trajectory. This multi-stage decoding from EEG signal to sEMG signal and then to the knee joint motion trajectory is achieved using two concatenated dendritic neural networks. Hierarchical signal processing establishes an indirect mapping relationship between the EEG signal and the knee joint motion trajectory, solving the technical problem of how to decode the knee joint motion trajectory using EEG and sEMG signals through concatenated fusion. This ensures that the two types of signals function sequentially during the decoding process, improving the targeting and effectiveness of knee joint motion trajectory decoding.

[0037] Example 2 This embodiment provides a specific implementation method for gait trajectory prediction based on dendritic neural networks. The operation method of this step is as follows.

[0038] Step 1: EEG, sEMG signals of movement-related muscle groups, and joint motion trajectory data were collected from the subjects while they walked on a treadmill (2.0 km / h). The muscles most relevant to lower limb gait were selected for calibration, and their sEMG signals were collected. The muscle channels used in this invention included the biceps brachii, vastus medialis, gastrocnemius, and tibialis anterior muscles of both legs. Twenty-one channels of EEG signals and one channel of electrooculography (EOG) signals were collected. The acquisition equipment used was a Brain Products 32Ch LiveAmp EEG cap and a wireless LiveAmp amplifier. The VICON T40S device and the wireless LiveAmp amplifier were synchronized, and the VICON T40S was used to collect the motion trajectories of 16 joints in the lower limbs. The sampling frequency for EEG signals was 500 Hz, and the sampling frequency for joint motion trajectory data was 100 Hz.

[0039] Step two involves preprocessing the acquired EEG, sEMG signals, and joint motion trajectory data by resampling (at a frequency of 1000Hz). The EEG signals undergo denoising processes such as rereference, bandpass filtering, removal of independent components related to electrooculography (EOG), and removal of motion artifacts. The sEMG signals undergo preprocessing including bandpass filtering and wavelet denoising. EEG and sEMG data are preprocessed in MATLAB, and the acquired joint motion trajectory data is processed using Nexus 2.6 software.

[0040] Step 3: Normalize the preprocessed EEG, sEMG, and joint motion trajectory signals. Specifically, calculate the mean and standard deviation of each channel in MATLAB, and then normalize each brain muscle electromyography channel using the mean and standard deviation.

[0041] Step four involves designing the structure of the dendritic neural network, including the number of samples in a batch, the number of network layers, etc., and training the network. First, based on the relationship between electromyography (EMG) signals and joint motion trajectories, appropriate dendritic neural network parameters are designed to construct a precise mapping relationship between the fused EMG signals and joint motion trajectories. The EMG signals are then divided into training and testing sets, and the constructed dendritic neural network model is trained.

[0042] Step 5: Use the preprocessed EEG and sEMG signals from a total of 29 channels as input to the dendritic neural network, and use the knee joint motion trajectory as the output of the dendritic neural network. For example... Figure 2 As shown, there are two ways to fuse EEG and sEMG signals. The first way is parallel fusion, which means that the 29 channels of EEG and sEMG signals are simultaneously input into the dendritic neural network, labeled as the knee joint motion trajectory, and the output of the network is the decoded knee joint motion trajectory. The second way is serial fusion. In this fusion method, two dendritic neural networks need to be trained. The input of the first dendritic neural network is the 21 channels of EEG signal, labeled as the sEMG signal, and the output is the decoded sEMG signal. The input of the second dendritic neural network is the output of the first dendritic neural network, labeled as the knee joint motion trajectory, and the output is the decoded knee joint motion trajectory.

[0043] After obtaining the decoded data, the Pearson correlation coefficient R and root mean square error (RMSE) between the decoded knee joint estimate and the actual trajectory were calculated. The decoding results show that when EEG and sEMG are decoded in parallel, the average Pearson correlation coefficient R is 0.97, and the RMSE is 0.11. The decoding rate is lower than in existing studies, and the RMSE is significantly lower, indicating that the decoded knee joint motion trajectory is not only more similar in shape but also more accurate in amplitude. In the series decoding of EEG and sEMG, in the first stage of decoding, i.e., decoding the sEMG signal from the EEG, a portion of the sEMG waveform can be decoded. Although the correlation between the decoded sEMG waveform and the measured waveform is not high (R=0.6), it is better than current studies decoding lower limb joint kinematic parameters from EEG (R<0.6). In the second stage, in the study of decoding the ankle joint trajectory from the obtained 8-channel sEMG, the highest Pearson correlation coefficient R=-0.47 between the decoded trajectory and the measured ankle joint trajectory was obtained.

[0044] Example 3 This embodiment provides a gait trajectory prediction system based on dendritic neural networks, which is used to implement the gait trajectory prediction methods based on dendritic neural networks in Embodiments 1 and 2. The system mainly includes: The motion data acquisition module is used to acquire sEMG and EEG signals of muscle groups during leg movements; The data preprocessing module is used to preprocess sEMG and EEG signals; The gait trajectory prediction module is used to input the preprocessed sEMG and EEG signals into the gait trajectory prediction model to predict the gait trajectory of both legs and obtain the knee joint movement trajectory of both legs. The gait trajectory prediction model uses a dendritic neural network model. The dendritic neural network model fuses the input sEMG signal and EEG signal, decodes the fused data, and obtains the knee joint movement trajectory of both legs based on the decoded data.

[0045] Example 4 In another embodiment of the present invention, a computer-readable storage medium is provided as a storage component within a terminal device, the function of which is to store programs and data. It should be noted that the computer-readable storage medium here encompasses not only the built-in storage components of the terminal device but also extended storage components supported by the device. Essentially, it is a tangible medium capable of containing or storing programs that can be invoked by or in conjunction with an instruction execution system, device, or apparatus. This storage medium provides storage areas for the terminal's operating system and stores one or more instructions suitable for processor loading and execution, which can constitute one or more computer programs containing program code.

[0046] Specifically, examples of computer-readable storage media (a non-exclusive list) include: electrical connections with one or more wires, portable disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable optical disc read-only memory, optical storage devices, magnetic storage devices, or any reasonable combination of the above types.

[0047] The storage medium may also include data signals propagated as part of a baseband portion or a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any reasonable combination of both. Furthermore, computer-readable storage medium may also refer to other readable media besides conventional readable storage media, capable of sending, propagating, or transmitting programs for use or operation by an instruction execution system, apparatus, or device. Program code on the storage medium can be transmitted via any suitable medium, including but not limited to wireless, wired, optical fiber, or any reasonable combination thereof.

[0048] The program code used to implement the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C. The execution modes of the program code include: running entirely on the user's computing device, running partially on the user's device as a standalone software package, running partially in a distributed manner on both the user's device and a remote computing device, or running entirely on a remote computing device or server. When a remote computing device is involved, the device can be connected to the user's computing device via any type of network such as a local area network (LAN) or a wide area network (WAN), or connected to an external computing device via the Internet through an Internet service provider.

[0049] The processor is capable of loading and executing one or more instructions stored in a computer-readable storage medium to implement the corresponding steps of the gait trajectory prediction method based on dendritic neural networks described in Embodiment 1.

[0050] Example 5 Figure 3 This is a schematic diagram of a computer device provided according to an embodiment of the present invention.

[0051] Please see Figure 3 The terminal device is a computer device. In this embodiment, the computer device 60 includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When executed by the processor 61, the computer program 63 implements the gait trajectory prediction method based on a dendritic neural network as described in this embodiment. To avoid repetition, these details are not elaborated here. Alternatively, when executed by the processor 61, the computer program 63 implements the functions of each model / unit in the computing system of Embodiment 1. To avoid repetition, these details are not elaborated here.

[0052] Computer device 60 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. Computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will understand that... Figure 3 This is merely an example of computer device 60 and does not constitute a limitation on computer device 60. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer device may also include input / output devices, network access devices, buses, etc.

[0053] The processor 61 may be a central processing unit (CPU), or other general-purpose processors, CPUs, graphics processing units (GPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, quantum computing-based data processing logic units, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0054] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or RAM of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device 60.

[0055] Furthermore, the memory 62 may include both internal storage units of the computer device 60 and external storage devices. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0056] Any references to memory, databases, or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0057] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

Claims

1. A gait trajectory prediction method based on dendritic neural networks, characterized in that, include: Acquire sEMG signals of muscle groups, kinematic signals of major joints of the lower limbs, and EEG signals of the cerebral cortex during leg movements; The sEMG signal, kinematic signal, and cerebral cortex EEG signal are preprocessed; The preprocessed sEMG signal, kinematic signal and EEG signal are input into the gait trajectory prediction model to predict the gait trajectory of both legs and obtain the knee joint movement trajectory of both legs. The gait trajectory prediction model uses a dendritic neural network model. The dendritic neural network model fuses the input sEMG signal and EEG signal, decodes the fused data, and obtains the knee joint movement trajectory of both legs based on the decoded data.

2. The gait trajectory prediction method based on dendritic neural networks according to claim 1, characterized in that, The sEMG and EEG signals are preprocessed, and the preprocessing process includes: After resampling the EEG signal, a denoising operation is performed based on the sampled EEG signal. The denoising operation includes at least re-reference, bandpass filtering, removal of independent components related to electrooculography, and removal of motion artifacts in sequence on the sampled EEG signal. The sEMG signal is subjected to bandpass filtering and wavelet denoising.

3. The gait trajectory prediction method based on dendritic neural networks according to claim 2, characterized in that, The bandpass filter used for bandpass filtering of the sEMG signal is a two-stage IIR filter based on Butterworth filter design, with a filtering frequency of 1-3Hz delta band.

4. The gait trajectory prediction method based on dendritic neural networks according to claim 2, characterized in that, It also includes normalizing the preprocessed sEMG and EEG signals.

5. The gait trajectory prediction method based on dendritic neural networks according to claim 1, characterized in that, The dendritic neural network model is used to simulate the dendritic structure of neurons, and the prediction method of the dendritic neural network model is as follows: In the formula, For the first The output of a dendritic branch layer For the first The output of a dendritic branch layer For the first From the first dendritic branch layer to the... The weight matrix of each dendritic branch layer For Hadama accumulation.

6. The gait trajectory prediction method based on dendritic neural networks according to claim 1 or 5, characterized in that, The dendritic neural network model includes a data fusion module, which performs parallel data fusion on the input sEMG and EEG signals; the parallel data fusion includes: When sEMG and EEG signals are simultaneously input into a dendritic neural network model, and the label in the dendritic neural network is set to the knee joint motion trajectory, the dendritic neural network can directly learn to decode the knee joint motion trajectory from the sEMG and EEG signals.

7. The gait trajectory prediction method based on dendritic neural networks according to claim 1 or 5, characterized in that, The data fusion module is used to perform concatenated data fusion on the input sEMG signal and EEG signal; the concatenated data fusion includes: Two cascaded dendritic neural networks are set up. The EEG signal is input into the first dendritic neural network, and the label of the first dendritic neural network is set as the sEMG signal. The sEMG signal is obtained by decoding based on the first dendritic neural network. The decoded sEMG signal is used as the input of the second dendritic neural network, and the label of the second dendritic neural network is set as the knee joint motion trajectory. The knee joint motion trajectory is output based on decoding the sEMG signal.

8. A gait trajectory prediction system based on a dendritic neural network, characterized in that, include: The motion data acquisition module is used to acquire sEMG and EEG signals of muscle groups during leg movements; The data preprocessing module is used to preprocess the sEMG and EEG signals; The gait trajectory prediction module is used to input the preprocessed sEMG and EEG signals into the gait trajectory prediction model to predict the gait trajectory of both legs and obtain the knee joint movement trajectory of both legs. The gait trajectory prediction model uses a dendritic neural network model. The dendritic neural network model fuses the input sEMG signal and EEG signal, decodes the fused data, and obtains the knee joint movement trajectory of both legs based on the decoded data.

9. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform the gait trajectory prediction method based on a dendritic neural network as described in any one of claims 1 to 7.

10. A computing device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including steps for performing the gait trajectory prediction method based on any one of claims 1 to 7.

Images