A continuous control method for a robotic arm based on bioelectrical signal recognition

Abstract

The present invention relates to the technical field of robotic arm control, and discloses a continuous control method for a robotic arm based on bioelectrical signal recognition. The method includes: collecting the electromyographic signals of the arm muscle groups and performing filtering and denoising processing on the electromyographic signals; using an association mapping relationship model to convert the filtered and denoised electromyographic signals into robotic arm control parameters, performing interpolation optimization based on control continuity on the robotic arm control parameters and converting them into joint torque parameters and end-effector torque parameters, and performing robotic arm control. The present invention uses the association mapping relationship model to extract features from the filtered and denoised electromyographic signals, extracts time-domain features and frequency-domain features respectively, maps the features representing the electromyographic signal pattern and the features representing the signal amplitude into robotic arm control parameters, and performs non-linear optimization on the robotic arm control parameters based on the error value sequence after interpolation optimization to obtain a more continuous and smooth optimization result, realizing the continuous control of the robotic arm based on bioelectrical signal recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of robotic arm control, and particularly to a continuous control method for a robotic arm based on bioelectrical signal recognition. Background Art

[0002] With the development of robot technology, artificial intelligence, and bioelectrical signal processing technology, controlling a robotic arm based on human bioelectrical signals (such as electroencephalogram signals, electromyogram signals, etc.) has become a research hotspot in the field of robot control in recent years. In particular, electromyogram signals (EMG), which are electrical signals generated by muscle activities and can reflect human motion intentions. The detection of EMG signals does not rely on traditional hardware interfaces and can be conveniently obtained through surface electrodes. For patients with limb dysfunction, it can achieve a more natural and intuitive control method. As an important part of human auxiliary equipment, robotic arms have broad application prospects in the fields of medical rehabilitation, disability assistance technology, industrial automation, etc. However, traditional EMG signal control methods often neglect the need for force control. Different tasks often require different levels of force control. Especially in some fine operations, the magnitude of the user's action force needs to be accurately identified and adjusted. Therefore, how to achieve force adaptive control based on EMG signals, enabling the robotic arm to automatically adjust the control force according to different user needs and operating scenarios, is an urgent problem to be solved in current research. Summary of the Invention

[0003] In view of this, the present invention provides a continuous control method for a robotic arm based on bioelectrical signal recognition, which accurately identifies the force requirements of the user in a specific operation by real-time sensing of the user's muscle activities, and realizes continuous control of the robotic arm.

[0004] To achieve the above object, a continuous control method for a robotic arm based on bioelectrical signal recognition provided by the present invention includes the following steps:

[0005] S1: Deploy surface electromyogram sensors on the arm muscle groups connected to the robotic arm, collect the electromyogram signals of the arm muscle groups, and perform filtering and denoising processing on the electromyogram signals;

[0006] S2: Construct an association mapping relationship model between the filtered and denoised electromyogram signals and the robotic arm control parameters;

[0007] S3: Use the association mapping relationship model to convert the filtered and denoised electromyogram signals into robotic arm control parameters, and perform interpolation optimization on the robotic arm control parameters based on control continuity;

[0008] S4: Convert the optimized robotic arm control parameters after interpolation into joint torque parameters and end-effector torque parameters, and send them to the control system of the robotic arm to adjust the torques of the robotic arm joints and the end effector, and control the rotations of the robotic arm joints and the end effector.

[0009] As a further improvement method of the present invention:

[0010] Optionally, the arm muscle groups include three muscle groups, namely the upper arm muscle group, the forearm muscle group, and the shoulder muscle group in sequence. Surface electromyogram sensors are respectively deployed on the three muscle groups to collect the electromyogram signals of the three muscle groups, and filter and denoise the electromyogram signals to remove the noise information in the electromyogram signals;

[0011] The representation form of the electromyogram signal is:

[0012] ;

[0013] Where:

[0014] represents the electromyogram signal of the i-th muscle group, represents the electromyogram signal the signal value at the n-th signal moment in, , represents the number of signal moments of the electromyogram signal, and the signal acquisition time interval of the electromyogram signal is .

[0015] Optionally, the filter and denoise processing flow is:

[0016] Use the discrete wavelet transform method to decompose the electromyogram signal into L-layer low-frequency wavelet coefficients and high-frequency wavelet coefficients. The discrete wavelet transform method filters the electromyogram signal using a low-pass filter and a high-pass filter to obtain the low-frequency wavelet coefficients and high-frequency wavelet coefficients of the first layer in sequence, and uses the low-pass filter and the high-pass filter to filter each layer of low-frequency wavelet coefficients to obtain the low-frequency wavelet coefficients and high-frequency wavelet coefficients of the next layer;

[0017] Retain the L-layer low-frequency wavelet coefficients, and perform filter and denoise processing on the high-frequency wavelet coefficients. The filter and denoise processing formula of the high-frequency wavelet coefficients is:

[0018] ;

[0019] ;

[0020] Where:

[0021] represents the high-frequency wavelet coefficient The filtering and denoising processing result;

[0022] represents the mean value of the high-frequency wavelet coefficients , represents the standard deviation of the high-frequency wavelet coefficients , represents the number of layers of the high-frequency wavelet coefficients ;

[0023] L represents the preset number of wavelet decomposition layers;

[0024] represents the adaptive threshold of the high-frequency wavelet coefficients ;

[0025] Perform signal reconstruction on the Lth layer of low-frequency wavelet coefficients and the high-frequency wavelet coefficients after L-layer filtering and denoising to obtain the electromyogram signal after filtering and denoising;

[0026] The electromyogram signal The corresponding electromyogram signal after filtering and denoising is , , is the signal value at the nth signal moment in the electromyogram signal after filtering and denoising .

[0027] Optionally, the feature extraction module is used to extract the biological features in the electromyogram signal after filtering and denoising, and convert the biological features into a biological feature matrix. The biological features include the time-domain features and frequency-domain features of the electromyogram signal after filtering and denoising. The feature mapping module is a support vector machine structure, which is used to extract the feature vector representing the moving direction of the robotic arm as the movement parameter vector and the feature vector representing the force at the end of the robotic arm as the force parameter vector from the biological feature matrix, and perform high-dimensional mapping on the movement parameter vector and the force parameter vector to obtain the robotic arm control parameters corresponding to the electromyogram signal after filtering and denoising.

[0028] Optionally, the time-domain features of the electromyogram signal after filtering and denoising include the root mean square value, average absolute value, and zero-crossing rate of the electromyogram signal after filtering and denoising. The root mean square value is the root mean square of the electromyogram signal after filtering and denoising, the average absolute value is the mean of the absolute values of the signal values in the electromyogram signal after filtering and denoising, and the zero-crossing rate is the number of times the electromyogram signal after filtering and denoising crosses the zero point;

[0029] The frequency-domain features of the electromyogram signal after filtering and denoising include the mean frequency, median frequency index, and total spectral energy of the electromyogram signal after filtering and denoising. The flow of extracting the frequency-domain features of the electromyogram signal after filtering and denoising is as follows:

[0030] Perform a Fourier transform on the filtered and denoised EMG signal to obtain the spectrum of the filtered and denoised EMG signal at N consecutive frequency indices. Calculate the center frequency of the N spectra as the mean frequency of the filtered and denoised EMG signal, calculate the total energy of the N spectra as the total spectral energy of the filtered and denoised EMG signal, and divide the N spectra into two continuously indexed frequency parts based on the median frequency index such that the spectral energies of the two parts are equal.

[0031] Optionally, use the feature extraction module to extract the biometric features of the filtered and denoised EMG signal and convert them into a biometric feature matrix. The representation form of the biometric feature matrix is :

[0032] ;

[0033] where:

[0034] T represents the transpose, represents the time-domain features of the filtered and denoised EMG signals of 3 muscle groups, represents the frequency-domain features of the filtered and denoised EMG signals of 3 muscle groups;

[0035] , , are the root mean square value, average absolute value, and zero crossing rate of the filtered and denoised EMG signal in sequence, are the mean frequency, median frequency index, and total spectral energy of the filtered and denoised EMG signal in sequence;

[0036] Use the feature mapping module to extract the feature vector representing the movement direction of the robotic arm from the biometric feature matrix F as the movement parameter vector :

[0037] ;

[0038] Use the feature mapping module to extract the feature vector representing the force at the end of the robotic arm from the biometric feature matrix F as the force parameter vector :

[0039] ;

[0040] Perform a high-dimensional mapping on the movement parameter vector and the force parameter vector to obtain the robotic arm control parameter G corresponding to the filtered and denoised EMG signal:

[0041] ;

[0042] ;

[0043] ;

[0044] Wherein:

[0045] is the control parameter for the moving direction of the robotic arm, is the control parameter for the force at the end of the robotic arm;

[0046] is the first weight matrix, is the second weight matrix, represents the first offset, represents the second offset, represents a convolution operation;

[0047] Obtain the control parameter sequence of the robotic arm control parameter G, and the control parameter sequence is :

[0048] ;

[0049] Wherein:

[0050] represents the interpolation optimization amount of the robotic arm control parameter at the signal acquisition time time - The interpolation optimization amount is the change amount of the robotic arm control parameter before and after interpolation optimization, and time represents the signal acquisition time of the myoelectric signal of, represents the interpolation optimization amount of the robotic arm control parameter at the signal acquisition time time - , , and M represents the length of the control parameter sequence.

[0051] Optionally, based on the control parameter sequence, perform interpolation optimization on the robotic arm control parameter G based on control continuity, and the interpolation optimization formula is:

[0052] ;

[0053] Wherein:

[0054] represents the interpolation optimization result of the robotic arm control parameter G, is the mean value of the control parameter sequence E, represents the standard deviation of the control parameter sequence E;

[0055] is the optimization amplitude control parameter, represents the optimization sensitivity control parameter, represents a preset adjustment parameter.

[0056] Optionally, the end torque parameter includes the position of the end effector of the robotic arm and the velocities in all directions; the conversion process of the joint torque parameter is as follows:

[0057] Obtain the end torque parameter, where the end torque parameter is the robotic arm control parameter after interpolation optimization , is the robotic arm movement direction control parameter after interpolation optimization, is the robotic arm end force control parameter after interpolation optimization;

[0058] The number of joints of the robotic arm is R, where the joint angle of the r-th joint is , , based on the joint angle and , based on the dynamic principle of the robotic arm, construct the Jacobian matrix J for controlling the torque of the robotic arm. The Jacobian matrix J is in the form of a matrix with 1 row and R columns, and the value of the r-th column element in the Jacobian matrix J is the partial derivative of the position of the end effector of the robotic arm with respect to the joint angle ;

[0059] Use the Jacobian matrix J to map the end torque parameter to the joint torque to obtain the joint torque parameter S:

[0060] ;

[0061] The joint torque parameter S is the velocities of the R joints of the robotic arm in all directions.

[0062] To solve the above problems, the present invention provides an electronic device, which includes:

[0063] A memory that stores at least one instruction;

[0064] A communication interface that enables the electronic device to communicate; and

[0065] A processor that executes the instructions stored in the memory to implement the above-mentioned robotic arm continuous control method based on bioelectric signal recognition.

[0066] To solve the above problems, the present invention further provides a computer-readable storage medium, in which at least one instruction is stored, and the at least one instruction is executed by a processor in an electronic device to implement the above-mentioned robotic arm continuous control method based on bioelectric signal recognition.

[0067] Compared with the prior art, the present invention proposes a robotic arm continuous control method based on bioelectric signal recognition, and this technology has the following advantages:

[0068] First, this solution proposes a method for generating robotic arm control parameters. It extracts the electromyogram (EMG) signals of muscle groups in different parts of the arm to represent bioelectric signals in multiple directions. A filtering and denoising process that combines an adaptive threshold is used to filter the EMG signals, adapting to the noise levels and signal characteristics of different muscle groups, effectively removing noise without losing the main components of the signal, providing a more accurate denoising effect without the need for manual threshold setting. A correlation mapping relationship model is used to extract features from the filtered and denoised EMG signals, extracting time-domain features and frequency-domain features respectively. The features representing the EMG signal patterns are used to map and obtain the robotic arm movement direction control parameters, and the features representing the signal amplitude are mapped to obtain the robotic arm end force control parameters, realizing the mapping of robotic arm control parameters based on bioelectric signal recognition.

[0069] Meanwhile, this solution proposes a method for continuous control of the robotic arm. Combining the interpolation optimization results of the robotic arm control parameters corresponding to past EMG signals, based on the error values of the robotic arm control parameters before and after interpolation optimization, the robotic arm control parameters are nonlinearly optimized to obtain a more continuous and smooth optimization result. During the optimization process, the amplitude and sensitivity are controlled to avoid excessive control of the optimization result by the error values. When the input value is large, the output of the tanh function tends to be stable, which can naturally suppress the influence of large-amplitude noise and improve the robustness of the robotic arm control parameters after interpolation optimization. The robotic arm control parameters after interpolation optimization are converted into joint torque parameters and end torque parameters, and sent to the control system of the robotic arm to control the movement speeds of the robotic arm joints and end effectors in different directions, realizing the continuous control of the robotic arm based on bioelectric signal recognition. Brief Description of the Drawings

[0070] Figure 1 It is a schematic flowchart of a method for continuous control of a robotic arm based on bioelectric signal recognition provided by an embodiment of the present invention.

[0071] The realization, functional features, and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed Embodiments

[0072] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0073] An embodiment of the present application provides a method for continuously controlling a robotic arm based on bioelectrical signal recognition. The execution subject of the method for continuously controlling a robotic arm based on bioelectrical signal recognition includes, but is not limited to, at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided in the embodiment of the present application. In other words, the method for continuously controlling a robotic arm based on bioelectrical signal recognition can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to: a single server, a server cluster, a cloud server, or a cloud server cluster, etc.

[0074] Referring to Figure 1 , Embodiment 1 of the present invention is:

[0075] A method for continuously controlling a robotic arm based on bioelectrical signal recognition, comprising the following steps:

[0076] S1: Deploy surface electromyography sensors on the arm muscle groups connected to the robotic arm, collect the electromyography signals of the arm muscle groups, and perform filtering and denoising processing on the electromyography signals.

[0077] The arm muscle groups include three muscle groups, and the first to third muscle groups are the upper arm muscle group, the forearm muscle group, and the shoulder muscle group in sequence. Surface electromyography sensors are respectively deployed on the three muscle groups to collect the electromyography signals of the three muscle groups, and filtering and denoising processing is performed on the electromyography signals to remove the noise information in the electromyography signals;

[0078] The representation form of the electromyography signal is:

[0079] ;

[0080] Where:

[0081] represents the electromyography signal of the i-th muscle group, represents the electromyography signal the signal value at the n-th signal moment in, , represents the number of signal moments of the electromyography signal, and the signal acquisition time interval of the electromyography signal is . In an embodiment of the present invention, the collected electromyography signal is used to control the robotic arm, and after time, the electromyography signal is collected again.

[0082] The filtering and denoising processing flow is:

[0083] The electromyogram signal is decomposed into L-layer low-frequency wavelet coefficients and high-frequency wavelet coefficients by using the discrete wavelet transform method. The discrete wavelet transform method filters the electromyogram signal by using a low-pass filter and a high-pass filter to sequentially obtain the low-frequency wavelet coefficients and high-frequency wavelet coefficients of the first layer, and uses the low-pass filter and the high-pass filter to filter each layer of low-frequency wavelet coefficients to obtain the low-frequency wavelet coefficients and high-frequency wavelet coefficients of the next layer;

[0084] Retain the L-layer low-frequency wavelet coefficients, and perform filtering and denoising processing on the high-frequency wavelet coefficients. The filtering and denoising processing formula of the high-frequency wavelet coefficients is:

[0085] ;

[0086] ;

[0087] Where:

[0088] represents the filtering and denoising processing result of the high-frequency wavelet coefficient ;

[0089] represents the mean value of the high-frequency wavelet coefficient ; represents the standard deviation of the high-frequency wavelet coefficient ; represents the number of layers of the high-frequency wavelet coefficient ;

[0090] L represents the preset number of wavelet decomposition layers;

[0091] represents the adaptive threshold of the high-frequency wavelet coefficient ;

[0092] Perform signal reconstruction on the L-layer low-frequency wavelet coefficients and the L-layer high-frequency wavelet coefficients after filtering and denoising to obtain the electromyogram signal after filtering and denoising;

[0093] The electromyogram signal corresponding to the electromyogram signal after filtering and denoising is , , is the signal value at the nth signal moment of the electromyogram signal after filtering and denoising. In a specific embodiment of the present invention, the signal reconstruction method is to superimpose the L-layer low-frequency wavelet coefficients and the L-layer high-frequency wavelet coefficients after filtering and denoising as the signal reconstruction result.

[0094] S2: Construct an association mapping relationship model between the electromyogram signal after filtering and denoising and the manipulator control parameters.

[0095] The feature extraction module is used to extract the biometric features in the EMG signal after filtering and denoising, and convert the biometric features into a biometric feature matrix. The biometric features include the time-domain features and frequency-domain features of the EMG signal after filtering and denoising. The feature mapping module is a support vector machine structure, which is used to extract the feature vector representing the moving direction of the robotic arm as the movement parameter vector and the feature vector representing the force at the end of the robotic arm as the force parameter vector from the biometric feature matrix, and perform high-dimensional mapping on the movement parameter vector and the force parameter vector to obtain the robotic arm control parameters corresponding to the EMG signal after filtering and denoising.

[0096] The time-domain features of the EMG signal after filtering and denoising include the root mean square value, average absolute value, and zero-crossing rate of the EMG signal after filtering and denoising. The root mean square value is the root mean square of the EMG signal after filtering and denoising, the average absolute value is the mean of the absolute values of the signal values in the EMG signal after filtering and denoising, and the zero-crossing rate is the number of times the EMG signal after filtering and denoising crosses the zero point.

[0097] The frequency-domain features of the EMG signal after filtering and denoising include the mean frequency, median frequency index, and total spectral energy of the EMG signal after filtering and denoising. The process for extracting the frequency-domain features of the EMG signal after filtering and denoising is as follows:

[0098] Perform Fourier transform on the EMG signal after filtering and denoising to obtain the spectrum of the EMG signal after filtering and denoising at N consecutive frequency indices. Calculate the central frequency of the N spectra as the mean frequency of the EMG signal after filtering and denoising, calculate the total energy of the N spectra as the total spectral energy of the EMG signal after filtering and denoising, and divide the N spectra into two parts with continuous frequency indices based on the median frequency index such that the spectral energies of the two parts are equal. The N consecutive frequency indices are from 0 to N - 1, where the median frequency index is between 0 and N - 1, and each frequency index corresponds to a spectrum and a frequency component. The frequency component corresponding to the jth frequency index is :

[0099] ;

[0100] Where:

[0101] ; represents the time interval between adjacent signal times in the EMG signal after filtering and denoising;

[0102] The EMG signal after filtering and denoising The mean frequency is :

[0103] ;

[0104] Where:

[0105] Denote the EMG signal after filtering and denoising The spectrum at the j-th frequency index.

[0106] S3: Use the association mapping relationship model to convert the EMG signal after filtering and denoising into robotic arm control parameters, and perform interpolation optimization on the robotic arm control parameters based on control continuity.

[0107] Use the feature extraction module to extract the biometric features of the EMG signal after filtering and denoising, and convert them into a biometric feature matrix. The representation form of the biometric feature matrix is :[[]]

[0108] ;

[0109] Where:

[0110] T represents the transpose, Denote the time-domain features of the EMG signal after filtering and denoising of 3 muscle groups, Denote the frequency-domain features of the EMG signal after filtering and denoising of 3 muscle groups;

[0111] , , Are the root mean square value, average absolute value, and zero crossing rate of the EMG signal after filtering and denoising in sequence, Are the mean frequency, median frequency index, and total spectral energy of the EMG signal after filtering and denoising in sequence;

[0112] Use the feature mapping module to extract the feature vector representing the moving direction of the robotic arm from the biometric feature matrix F as the moving parameter vector :[[]]

[0113] ;

[0114] Use the feature mapping module to extract the feature vector representing the force at the end of the robotic arm from the biometric feature matrix F as the force parameter vector :[[]]

[0115] ;

[0116] Perform high-dimensional mapping on the moving parameter vector and the force parameter vector to obtain the robotic arm control parameter G corresponding to the EMG signal after filtering and denoising:

[0117] ;

[0118] ;

[0119] ;

[0120] Wherein:

[0121] is the control parameter for the moving direction of the robotic arm, is the control parameter for the force at the end of the robotic arm;

[0122] is the first weight matrix, is the second weight matrix, represents the first offset, represents the second offset, represents a convolution operation;

[0123] Obtain the control parameter sequence of the robotic arm control parameter G, and the control parameter sequence is :

[0124] ;

[0125] Wherein:

[0126] represents the interpolation optimization amount of the robotic arm control parameter at the signal acquisition time time - The interpolation optimization amount is the change amount of the robotic arm control parameter before and after interpolation optimization, and time represents the myoelectric signal The signal acquisition time of, represents the interpolation optimization amount of the robotic arm control parameter at the signal acquisition time time - and M represents the length of the control parameter sequence. In an embodiment of the present invention, the feature mapping module can be trained using one or a combination of the ADAM optimization algorithm, Newton's iterative method, and gradient descent method. Based on the control parameter sequence, perform interpolation optimization on the robotic arm control parameter G based on control continuity, where the interpolation optimization formula is:

[0127]

[0128] ;

[0129] Wherein:

[0130] represents the interpolation optimization result of the robotic arm control parameter G, is the mean value of the control parameter sequence E, represents the standard deviation of the control parameter sequence E;

[0131] is the optimization amplitude control parameter, Represents the optimized sensitivity control parameter, Represents the preset adjustment parameter.

[0132] S4: Convert the interpolated and optimized manipulator control parameters into joint torque parameters and end-effector torque parameters, and send them to the control system of the manipulator to adjust the torques of the manipulator joints and the end effector, and control the rotations of the manipulator joints and the end effector.

[0133] The end-effector torque parameters include the position of the end effector of the manipulator and the speeds in all directions; the conversion process of the joint torque parameters is as follows:

[0134] Obtain the end-effector torque parameters, where the end-effector torque parameters are the interpolated and optimized manipulator control parameters , is the interpolated and optimized manipulator movement direction control parameter, is the interpolated and optimized manipulator end-effector force control parameter;

[0135] The number of joints of the manipulator is R, where the joint angle of the r-th joint is , , based on the joint angle and , based on the dynamic principle of the manipulator, construct the Jacobian matrix J for controlling the manipulator torque. The Jacobian matrix J is in the matrix form of 1 row and R columns, and the value of the r-th column element in the Jacobian matrix J is the partial derivative of the position of the manipulator end effector with respect to the joint angle ;

[0136] Use the Jacobian matrix J to map the end-effector torque parameters to joint torques to obtain the joint torque parameters S:

[0137] ;

[0138] The joint torque parameter S is the speed of the R joints of the manipulator in all directions.

[0139] Embodiment 2:

[0140] This solution conducts a comparative experiment on the continuous control method of the manipulator based on bioelectric signal recognition, the PID control method, the fuzzy control method, and the reinforcement learning method. By setting a target task for the manipulator, evaluate the control performance of the manipulator under different control methods. The comparative experiment results are shown in Table 1:

[0141] Table 1

[0142] ;

[0143] As shown in Table 1, the robotic arm continuous control method based on bioelectrical signal recognition is significantly superior to other methods in terms of response time, error, control accuracy, and task completion rate. The interpolation optimization method based on bioelectrical signals and continuous control parameters can significantly reduce the response time and error and improve the control accuracy of the robotic arm.

[0144] It should be understood that the above embodiments are for illustrative purposes only and are not limited by this structure in the scope of the patent application.

[0145] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the superiority or inferiority of the embodiments. And the term "including" or "comprising" or any other variant thereof in this article is intended to cover a non-exclusive inclusion, so that a process, apparatus, article or method including a series of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, apparatus, article or method. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, apparatus, article or method including the element.

[0146] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above and includes several instructions for causing a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present invention.

[0147] The above are only the preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A method for continuous control of a robotic arm based on bioelectric signal recognition, characterized in that: The method comprises: S1: Deploy surface electromyography sensors on the arm muscle groups connected to the robotic arm to collect electromyography signals of the arm muscle groups and perform filtering and denoising on the electromyography signals; The robot arm is composed of connecting rods, joints, end effectors and control systems. The connecting rods are used to support the joints and the end effectors. The joints are used to realize the rotation and movement between the connecting rods. The end effectors are used to complete the components of the refined operation tasks. The types of end effectors include grippers, welding heads and cutting tools. S2: constructing a correlation mapping relationship model between the filtered and denoised electromyographic signal and the robotic arm control parameters, wherein the robotic arm control parameters include the robotic arm movement direction control parameters and the robotic arm end force control parameters, and the correlation mapping relationship model includes a feature extraction module and a feature mapping module; The robot arm movement direction control parameter is the position of the robot arm end effector, and the robot arm end force control parameter is the speed of the robot arm end effector in each direction; S3: The filtered and denoised electromyographic signals are converted into robot arm control parameters using the association mapping relationship model, and the robot arm control parameters are interpolated and optimized based on control continuity; S4: converting the interpolation optimized robot control parameters into joint torque parameters and end torque parameters, and sending them to the robot control system to adjust the torque of the robot joint and the end effector, and control the rotation of the robot joint and the end effector; The feature extraction module is used to extract the biological features of the electromyographic signal after filtering and denoising, and convert it into a biological feature matrix. The representation form of the biological feature matrix is : ； in: T stands for transpose, Represents the time domain characteristics of the EMG signals after filtering and denoising of the three muscle groups. The frequency domain characteristics of the EMG signals after filtering and denoising of the three muscle groups are shown; , , The electromyographic signals after filtering and denoising are The root mean square value, mean absolute value, and zero crossing rate of The electromyographic signals after filtering and denoising are The mean frequency, median frequency index and total spectrum energy of ; The feature mapping module is used to extract the feature vector representing the movement direction of the robot arm from the biological feature matrix F as the movement parameter vector : ； The feature mapping module is used to extract the feature vector representing the force of the end of the robot arm from the biological feature matrix F as the force parameter vector : ； For the movement parameter vector and the velocity parameter vector Perform high-dimensional mapping to obtain the robot arm control parameters G corresponding to the electromyographic signal after filtering and denoising: ； ； ； in: is the control parameter of the robot arm's moving direction, is the force control parameter of the end of the robot arm; is the first weight matrix, is the second weight matrix, represents the first offset, represents the second offset, Represents the convolution operation; Get the control parameter sequence of the robot arm control parameter G, the control parameter sequence is : ； in: Indicates that at the signal acquisition time time- The interpolation optimization amount of the robot arm control parameters is the change of the robot arm control parameters before and after the interpolation optimization, and time represents the electromyographic signal The signal acquisition time is Indicates that at the signal acquisition time time- The interpolation optimization amount of the robot arm control parameters, , M represents the length of the control parameter sequence; Based on the control parameter sequence, the robot control parameter G is interpolated and optimized based on control continuity, wherein the interpolation optimization formula is: ； in: represents the interpolation optimization result of the robot control parameter G, is the mean of the control parameter sequence E, represents the standard deviation of the control parameter sequence E; To optimize the amplitude control parameters, represents the optimization sensitivity control parameter, Indicates the preset adjustment parameters.

2. A method for continuous control of a robotic arm based on bioelectric signal recognition as claimed in claim 1, characterized in that: The arm muscle groups include three muscle groups, the first to third muscle groups are the upper arm muscle group, the forearm muscle group and the shoulder muscle group, and surface electromyography sensors are respectively deployed on the three muscle groups to collect electromyography signals of the three muscle groups, and the electromyography signals are filtered and denoised to remove noise information in the electromyography signals; The electromyographic signal is expressed as: ； in: represents the electromyographic signal of the i-th muscle group, Represents myoelectric signal The signal value at the nth signal moment in , , Represents the number of signal moments of the electromyographic signal. The signal acquisition time interval of the electromyographic signal is .

3. A method for continuous control of a robotic arm based on bioelectric signal recognition as claimed in claim 2, characterized in that: The filtering and denoising process is as follows: Decomposing the electromyographic signal into L layers of low-frequency wavelet coefficients and high-frequency wavelet coefficients by using a discrete wavelet transform method, wherein the discrete wavelet transform method uses a low-pass filter and a high-pass filter to filter the electromyographic signal, and sequentially obtains the low-frequency wavelet coefficients and high-frequency wavelet coefficients of the first layer, and uses a low-pass filter and a high-pass filter to filter the low-frequency wavelet coefficients of each layer, and obtains the low-frequency wavelet coefficients and high-frequency wavelet coefficients of the next layer; The low-frequency wavelet coefficients of the Lth layer are retained, and the high-frequency wavelet coefficients are filtered and denoised. The filtering and denoising formula is: ； ； in: Represents high-frequency wavelet coefficients The filtering and denoising results; Represents high-frequency wavelet coefficients The mean of Represents high-frequency wavelet coefficients The standard deviation of Represents high-frequency wavelet coefficients The number of layers; L represents the preset number of wavelet decomposition layers; Represents high-frequency wavelet coefficients Adaptive threshold of Reconstructing the L-th layer of low-frequency wavelet coefficients and the L-th layer of high-frequency wavelet coefficients after filtering and denoising to obtain a filtered and denoised electromyographic signal; The electromyographic signal The corresponding EMG signal after filtering and denoising is: , , The electromyographic signal after filtering and denoising The signal value at the nth signal moment in .

4. A method for continuous control of a robotic arm based on bioelectric signal recognition as claimed in claim 1, characterized in that: The time domain characteristics of the electromyographic signal after filtering and denoising include the root mean square value, average absolute value and zero crossing rate of the electromyographic signal after filtering and denoising, wherein the root mean square value is the root mean square of the electromyographic signal after filtering and denoising, the average absolute value is the mean of the absolute values ​​of the signal values ​​in the electromyographic signal after filtering and denoising, and the zero crossing rate is the number of times the electromyographic signal after filtering and denoising crosses zero points; The frequency domain features of the electromyographic signal after filtering and denoising include the mean frequency, median frequency index and total spectrum energy of the electromyographic signal after filtering and denoising. The frequency domain feature extraction process of the electromyographic signal after filtering and denoising is as follows: Perform Fourier transform on the filtered and denoised electromyographic signal to obtain the spectrum of the filtered and denoised electromyographic signal under N continuous frequency indexes, calculate the center frequency of the N spectrum as the mean frequency of the filtered and denoised electromyographic signal, calculate the total energy of the N spectrum as the total spectrum energy of the filtered and denoised electromyographic signal, and divide the N spectrum into two parts with continuous frequency indexes based on the median frequency index so that the spectrum energy of the two parts is equal.

5. The method for continuous control of a robotic arm based on bioelectric signal recognition according to claim 1, characterized in that: The end torque parameters include the position of the end effector of the robot arm and the speed in each direction; the conversion process of the joint torque parameters is: Get the end torque parameter, where the end torque parameter is the robot control parameter after interpolation optimization , is the interpolation optimized robot arm movement direction control parameter, The force control parameters of the end of the robot arm after interpolation optimization; The number of joints of the robot arm is R, and the joint angle of the rth joint is , , based on the joint angles and Based on the dynamic principle of the robot arm, the Jacobian matrix J for controlling the torque of the robot arm is constructed. The Jacobian matrix J is a matrix in the form of 1 row and R columns. The rth column element value in the Jacobian matrix J is the position of the robot end effector to the joint angle The partial derivative of The Jacobian matrix J is used to map the end torque parameter to the joint torque to obtain the joint torque parameter S: ； The joint torque parameter S is the speed of the R joints of the robot arm in each direction.

Images