A brain-controlled wheelchair method and system based on gradient paradigm and human-machine closed loop

Abstract

This invention discloses a brain-controlled wheelchair method and system based on a gradient paradigm and human-machine closed loop. The method includes: constructing a gradient-controlled brain feature evoked potential system based on the spatial distribution characteristics of steady-state visual evoked potentials to acquire feature potentials; constructing an EEG decoding system using a composite neural network to output EEG control commands; wherein the composite neural network is constructed and trained using the collected EEG data and the feature potentials; acquiring initial wheelchair control commands based on the EEG control commands and original wheelchair control commands, and generating final wheelchair control commands using a PID-sliding mode control method, thereby driving the wheelchair to perform corresponding movements. This application can solve the problems of poor robustness, large response error, and easy visual fatigue caused by discrete control in existing brain-controlled wheelchair technologies.

Description

Technical Field

[0001] This invention relates to the fields of brain-computer interface and wheelchair control technology, and in particular to a brain-controlled wheelchair method and system based on gradient paradigm and human-machine closed loop. Background Technology

[0002] Brain-controlled wheelchair technology is an important research area aimed at helping people with limited mobility achieve greater freedom of movement and independence. Traditional wheelchair control methods typically involve manual operation or verbal commands, but these methods may not be convenient or suitable for some people with physical disabilities. Therefore, the development of brain-controlled wheelchair technology is of great potential importance.

[0003] Current brain-controlled wheelchair technology is primarily based on brain-computer interface principles, utilizing the user's electroencephalogram (EEG) signals to control the wheelchair. EEG signals are bioelectrical signals acquired from the user's brain, providing information about the user's intentions and state. A typical brain-controlled wheelchair system includes EEG signal acquisition equipment, a signal processing module, decoding algorithms, and a wheelchair control unit.

[0004] In past research, EEG signal acquisition devices typically use EEG caps and amplifiers to acquire users' EEG data. This data undergoes preprocessing and filtering to reduce noise interference, and feature extraction is performed to capture information relevant to wheelchair control. Decoding of the EEG signals is usually achieved through neural networks or other machine learning algorithms, translating the user's EEG signals into wheelchair control commands.

[0005] Existing brain-controlled wheelchair systems face several challenges in practical applications. For example, improving the accuracy and stability of EEG signals, enhancing the user experience to reduce fatigue caused by visual evoked patterns, and achieving real-time closed-loop control are all issues that need to be addressed.

[0006] In addition, wheelchair control needs to take into account movements in multiple directions, including forward, backward, left turn, and right turn, thus requiring highly precise control algorithms. Summary of the Invention

[0007] Based on the technical problems existing in the prior art, the purpose of this application is to provide a brain-controlled wheelchair method and system based on gradient paradigm and human-machine closed loop, in order to solve the problems of poor robustness, large response error and easy visual fatigue caused by discrete control in the existing brain-controlled wheelchair technology.

[0008] On the one hand, to achieve the above objectives, this invention provides a brain-controlled wheelchair method based on a gradient paradigm and human-machine closed loop, comprising:

[0009] Based on the spatial distribution characteristics of steady-state visual evoked potentials, a gradient-controlled brain feature evoked system was constructed to obtain characteristic potentials.

[0010] A brainwave decoding system is constructed using a composite neural network to output brainwave control commands; wherein, the composite neural network is constructed and trained by collecting brainwave data and the characteristic potentials;

[0011] Based on the EEG control instructions and the original wheelchair control instructions, an initial wheelchair control instruction is obtained, and a final wheelchair control instruction is generated using the PID-sliding mode control method. The wheelchair is then driven to perform corresponding movements using the final wheelchair control instruction.

[0012] Preferably, the gradient-controlled brain feature induction system comprises:

[0013] By providing visual stimuli to the user through stimulus sources, the location of the user's gaze target point is obtained. The stimulus sources are stimulus patterns used to control the movement and rotation of the wheelchair and patterns used to control the proportion of brain control permissions.

[0014] Based on the area enclosed by the stimulus pattern used to control the movement and rotation of the wheelchair, i.e. the target area, a user gaze target point is obtained. A Cartesian two-dimensional coordinate system is set at the position of the gaze target point. The position of the gaze target point is described by the two-dimensional coordinates, and the two components of the two-dimensional coordinates are converted into the wheelchair's movement speed and rotation angular velocity, respectively. The stimulus pattern used to control the movement and rotation of the wheelchair changes in the form of a simple harmonic wave.

[0015] Based on the pattern connecting the line segments used to control the proportion of brain control permissions, the second user gaze target point is obtained, which is used to describe the weight of human brain commands.

[0016] Preferably, the composite neural network comprises:

[0017] Feature extraction sub-network: used for feature extraction;

[0018] EEG command output subnetwork: used to output EEG commands;

[0019] The feature extraction subnetwork includes a three-layer DenseNet network structure, and the EEG command output subnetwork includes a three-layer fully connected network structure.

[0020] Preferably, training the feature extraction sub-network includes:

[0021] Target points are evenly distributed in the target area, and users are asked to focus on each target point while their EEG data is collected. The EEG data is then preprocessed.

[0022] The ratio of the distance between the target point and each stimulus block is measured, and the ratio is normalized to the range of [0,1] to obtain the ideal EEG feature vector of the EEG signal induced by the target location, which is used for subsequent EEG command decoding.

[0023] Based on the evoked patterns of SSVEP in different visual positions of humans, a feature extraction sub-network structure that can link user EEG data and EEG feature vectors is constructed through supervised training.

[0024] Preferably, training the EEG command output subnetwork includes:

[0025] Establish a relationship function between the target location coordinates and the output brain control command, and obtain the ideal EEG feature vector and its corresponding output brain control command data by uniform sampling method;

[0026] The EEG command output subnetwork is trained using supervised training based on the ideal EEG feature vector and the corresponding brain control command data.

[0027] Preferably, the construction of the EEG decoding system includes:

[0028] Extracting characteristic potentials and locating the gaze target point;

[0029] The extraction of feature potentials includes: extracting the evoked intensities of several evoked potentials at different frequencies through the feature extraction subnetwork in the composite neural network to form a feature vector;

[0030] The process of locating the gaze target point includes: finding the maximum value among the components of the feature vector to determine the user's intention; removing the feature vector components that the user has no intention of controlling, and using the ratio of component values ​​to correspond to the ratio of distances between the annotation target point and each stimulus pattern to determine the location of the target point being gazed at; obtaining the location coordinates of the target point through calculation, and outputting the EEG control command.

[0031] The user's intention is to control the movement of the wheelchair or adjust the level of mind control permissions.

[0032] Preferably, the initial wheelchair control command is:

[0033] v target =(1+λ)v brain +(1-λ)v esc +v old

[0034] Among them, v target For the target speed that the wheelchair needs to achieve, v old v is the current speed of the wheelchair, λ is the brain control access coefficient, and v brain v is the speed generated by the current brain-controlled command.esc This is the reverse velocity generated when the wheelchair approaches the obstacle.

[0035] Preferably, the PID-sliding mode control method is as follows:

[0036]

[0037] Where u(t) represents the control command (including angular velocity and linear velocity) to be applied at the current moment; e(t) is the difference between the ideal value and the actual value at the current moment; e(·) is the error function; K p K i K d These are proportional, integral, and differential gains, respectively.

[0038] On the other hand, to achieve the above objectives, the present invention also provides a brain-controlled wheelchair system based on gradient paradigm and human-machine closed loop, comprising:

[0039] The system includes a human brain feature induction module, a data acquisition module, a data processing module, a brain feature recognition module, and a human-machine closed-loop control module.

[0040] The human brain feature induction module is used to construct a brain feature induction system based on stimulus sources;

[0041] The data acquisition module is used to collect the user's electroencephalogram (EEG) data;

[0042] The data processing module is used to preprocess the EEG data and construct and train a composite neural network;

[0043] The brain feature recognition module is used to construct an EEG decoding and classification system based on the composite neural network;

[0044] The human-machine closed-loop control module is used to construct a human-machine closed-loop control system based on the EEG control commands and the original wheelchair control commands, using the PID-sliding mode control method.

[0045] Preferably, the human-machine closed-loop control system includes:

[0046] The brain control command sending module is used to send brainwave control commands to the brain control command receiving module;

[0047] The human brain control command receiving module is used to receive transmitted brainwave control commands.

[0048] Wheelchair original control command output module, used to output original wheelchair commands;

[0049] The final wheelchair control command output module is used to comprehensively analyze the EEG control command and the original wheelchair command to generate the final wheelchair control command, and generate a control signal through the PID-sliding mode control algorithm to output to the wheelchair execution terminal.

[0050] The wheelchair actuator is used to receive the control signals and drive the wheelchair to perform corresponding movements.

[0051] Compared with the prior art, the present invention has the following advantages and technical effects:

[0052] This application can solve the problems of poor robustness, large response error and easy visual fatigue caused by the discrete control of existing brain-controlled wheelchair technology. Attached Figure Description

[0053] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0054] Figure 1 This is a flowchart of a brain-controlled wheelchair method based on gradient paradigm and human-machine closed loop according to an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram of the gradient-induced paradigm in Embodiment 1 of the present invention;

[0056] Figure 3 This is a schematic diagram of the coordinate positioning of the gaze target point provided in Embodiment 1 of the present invention;

[0057] Figure 4 This is a schematic diagram of the brain-controlled wheelchair command generation method in Embodiment 1 of the present invention when no obstacle is detected (left) and when an obstacle is detected (right);

[0058] Figure 5 This is a schematic diagram of the PID control method used in Embodiment 1 of the present invention;

[0059] Figure 6 This is a schematic diagram of the brain-controlled wheelchair system based on gradient paradigm and human-machine closed loop, which is a second embodiment of the present invention. Detailed Implementation

[0060] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0061] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0062] Example 1

[0063] like Figure 1 This is a schematic diagram of the brain-controlled wheelchair method based on gradient paradigm and human-machine closed loop proposed in this application, which includes the following steps:

[0064] S1. Based on the spatial distribution characteristics of steady-state visual evoked potentials, a gradient-controlled brain feature evoked system is constructed to obtain characteristic potentials;

[0065] The stimulation sources include: four stimulation patterns for controlling the movement and rotation of the wheelchair and two patterns for controlling the proportion of brain control authority. The brightness of the stimulation patterns changes in a simple harmonic form over time.

[0066] Four stimulus patterns used to control the wheelchair's movement and rotation enclose a quadrilateral area, which the user can focus on at any point as needed. This location is called gaze target point location one. A Cartesian two-dimensional coordinate system is set up in this area to describe gaze target point location one, and the two components of this coordinate system are translated into human brain control commands for the wheelchair's movement speed and rotation angular velocity, respectively.

[0067] Two icons used to control the proportion of brain-controlled permissions in the wheelchair connect to form a line segment, which the user can gaze at at any point as needed. This point is called gaze target point position two. It describes the weight of human brain commands in the human-machine closed-loop control system.

[0068] The location of the gaze target point identified by the system is fed back to the visual evoked interface in real time to provide real-time feedback to the user.

[0069] In this embodiment, six visual stimuli are used to construct a brain feature eliciting system. For example... Figure 2 As shown, the visual stimulus blocks are circular in shape, white in color, and flash at frequencies of 10Hz, 11Hz, 12Hz, 13Hz, 14Hz, and 15Hz.

[0070] The shaded area on the left represents the region the user wants to focus on when controlling the brain-controlled wheelchair. Previous research indicates that the intensity of the evoked SSVEP signal gradually decreases as the distance between the stimulus block and the user's visual center increases. By obtaining the SSVEP intensity ratios (10, 11, 12, and 13 Hz) of the four stimulus blocks on the left, the specific location of the shaded area the user is focusing on can be calculated. Figure 3 The coordinates shown record the travel speed and angular velocity corresponding to the target point.

[0071] like Figure 3 As shown, the vertical line connecting the two stimulation blocks on the right is the area that the user can focus on when they want to adjust the size of the brain control permissions.

[0072] In the interface, the gaze points of the two control parts, as determined by the system, will be reflected in the interface in real time (the position of the asterisk in the image), as visual feedback to the user.

[0073] S2. Collect EEG data and, based on the characteristic potentials, construct and train a composite neural network;

[0074] First, the collected EEG signals are preprocessed and sliced. The preprocessing process includes filtering, downsampling, removing bad segments, rereference, identifying and removing spurious components, etc.

[0075] Composite neural networks include:

[0076] Feature extraction sub-network: used for feature extraction;

[0077] EEG command output subnetwork: used to output EEG commands;

[0078] The feature extraction subnetwork consists of a three-layer DenseNet network structure, and the EEG command output subnetwork consists of a three-layer fully connected network structure.

[0079] Training a composite neural network includes training a feature extraction subnetwork and training an EEG command output subnetwork; the two training processes can be completed in parallel.

[0080] The composite network is trained using the collected EEG data. The feature extraction subnetwork needs to be trained separately for different users. Target points are set in the target region according to a uniform distribution. By measuring the ratio of the distance between the target point and four (two) stimulus blocks, the largest ratio is set as 1, which is used as the ideal EEG feature induced by the target point location. Numerous data points can be obtained from several target point locations for training the feature extraction subnetwork. Based on the evoked patterns of SSVEP in different visual field positions, a relational function between the sample target location and the brain control command output can be established according to user experience requirements. This generates a set of data for training the feature extraction subnetwork.

[0081] Specifically, in this embodiment, the input data for training the feature extraction subnetwork is a 1-second segment of a 3-channel EEG signal. The network output is a vector of length 6, representing the extracted EEG feature vector. The first four elements represent the features controlling wheelchair movement on the left, and the last two elements represent the features controlling the degree of brain control authority on the right. During training, when the target point is located in the left shaded area, the last two elements are 0; similarly, when the target point is located on the right straight line segment, the first four elements are 0. In training the output command subnetwork, the input data is a 6-length vector describing the EEG features, and the output data is a 3-dimensional vector describing brain control commands. The first two components represent the linear velocity and angular velocity in the brain-controlled wheelchair movement commands, respectively, and the last component represents the percentage of brain control authority. If the last two components in the input data vector are 0, this component is -1, indicating that the brain control authority remains unchanged; otherwise, it is limited to the range of 0 to 1 according to a linear distribution.

[0082] S3. Construct an EEG decoding system through the composite neural network and output EEG control commands;

[0083] Constructing an EEG decoding system, including:

[0084] Extracting characteristic potentials and locating the gaze target point;

[0085] Feature potential extraction involves extracting the evoked potential intensities of six different frequencies using a feature extraction subnetwork in a composite neural network to form a feature vector.

[0086] Locating the fixation target involves three steps: First, find the maximum value among the feature vector components to determine whether the user's intention is to control the wheelchair movement or adjust the brain control permissions; second, remove the feature vector components that the user has no intention to control, and match the ratio of component values ​​to the ratio of the distances between the annotated target point and each stimulus pattern to determine the location of the user's fixation target point; third, output the EEG control command based on the calculated target point location coordinates.

[0087] After training, a composite neural network can be used to build an EEG decoding system. By using a sliding time window method, the most recent 1 second of EEG signals acquired in real time is continuously input into the trained composite neural network system, which can output EEG control commands in real time.

[0088] S4. Based on the EEG control instructions and the original wheelchair control instructions, obtain the initial wheelchair control instructions, and generate the final wheelchair control instructions through the PID-sliding mode control method. Drive the wheelchair to perform corresponding movements through the final wheelchair control instructions.

[0089] The brainwave control commands include movement commands and rotation commands, which are determined by the horizontal and vertical coordinates of the identified gaze target point. When a brainwave control command is triggered, a brainwave decoding result is output, causing the brain-controlled wheelchair to generate a velocity vector Δv in the direction of movement.

[0090] It can be represented as: v brain =v brain,old +Δv

[0091] Among them, v brain v represents the speed at which the current mind control command is generated. brain,old This represents the speed at which the mind control command was generated at the previous moment.

[0092] The original wheelchair control command refers to an automatic obstacle avoidance measure that the wheelchair system itself will generate when obstacles appear in the surrounding environment during brain-controlled wheelchair operation. When the radar detects an obstacle, a reverse velocity is applied to the wheelchair itself, the magnitude of which increases as it gets closer to the obstacle, and can be expressed as: ||v esc ||=k(Ld);

[0093] Among them, v esc The reverse velocity generated when the wheelchair approaches the obstacle, L is the set obstacle avoidance judgment distance, d is the distance between the wheelchair and the obstacle, and k is a constant coefficient.

[0094] The final wheelchair control command, which is a combination of brain-controlled commands and the original wheelchair control commands, can be represented as: v target =(1+λ)v brain +(1-λ)v esc +v old ;

[0095] Among them, v target For the target speed that the wheelchair needs to achieve, v old Let λ be the current speed of the wheelchair, and λ be the brain control authority coefficient, which ranges from [-1, 1].

[0096] The PID-sliding mode control algorithm treats the velocity and angular velocity in each direction as independent control channels, designing independent PID controllers. Ideal control effects can be achieved through parameter tuning. Then, a sliding mode control correction term is introduced on top of the PID control to improve system robustness and reduce corresponding system errors. Sliding mode control can slide the system state to a stable hyperplane. The calculation formula for PID control is as follows:

[0097]

[0098] Where u(t) is the output of the control system, representing the control command (including angular velocity and linear velocity) to be applied at the current moment; e(t) is the difference between the ideal value and the actual value at the current moment; K p K i K d These are proportional, integral, and differential gains, respectively.

[0099] like Figure 4 As shown, the left side illustrates the control command generation method when no obstacle is detected, and the right side illustrates the control command generation method when an obstacle is detected. Referring to the brain control authority component in EEG control commands, the EEG control commands and the wheelchair are combined with the original wheelchair control commands. Finally, as shown... Figure 5 The PID-sliding mode control method is used to achieve fast, stable and accurate human-machine closed-loop control.

[0100] Example 2

[0101] like Figure 6 As shown, this application also provides a brain-controlled wheelchair system based on gradient paradigm and human-machine closed loop, including: a human brain feature induction module, a data acquisition module, a data processing module, a brain feature recognition module, and a human-machine closed loop control module;

[0102] The human brain feature induction module is used to construct a brain feature induction system based on stimulus sources;

[0103] In this embodiment, a brain feature eliciting system is constructed using a gradient-controllable stimulation paradigm. The stimulation source includes six circular flashing stimulation blocks, four of which are located on the left side, forming an area used to control the movement of the brain-controlled wheelchair, and the two on the right side are connected in a straight line segment to adjust the brain control authority level;

[0104] The human brain feature evoked module is also used to display visual stimulus sources, target areas, and the location of the identified focus point for visual feedback.

[0105] The data acquisition module is used to collect the user's brain feature map data;

[0106] The data acquisition module includes an EEG cap, an amplifier, and a data transmission interface; the EEG cap is used to acquire the user's EEG data; the amplifier is used to amplify the EEG data; and the data transmission interface is used to transmit the EEG data.

[0107] The data processing module is used to preprocess the acquired EEG data, including filtering, downsampling, removing bad segments, rereferencing, identifying and removing spurious components, etc. This improves the recognizability of EEG signals and reduces the learning cost of subsequent neural networks.

[0108] The brain feature recognition module is used to identify the location of the target point that the user is looking at based on the preprocessed EEG signal and convert it into brain control commands for the brain-controlled wheelchair.

[0109] In this embodiment, the subnetwork used to identify the location of the target point being gazed at by the user is a three-layer DenseNet network structure, which can extract EEG features. The network used to convert EEG features into brain control commands is a three-layer fully connected network structure.

[0110] The human-machine closed-loop control module is used to construct a human-machine closed-loop control system based on the PID-sliding mode control method, according to the EEG control commands and the original wheelchair control commands.

[0111] Specifically, it includes:

[0112] The human brain control command sending module is used to send the brain control commands output by the brain feature recognition module to the human brain control command receiving module;

[0113] The human brain control command receiving module is used to receive and transmit human brain control commands;

[0114] The wheelchair's original control command output module includes a detection radar and a detection radar signal analysis module. In this embodiment, a lidar is used to detect the environment around the wheelchair within a specified radius. When an obstacle is detected nearby, the obstacle point closest to the wheelchair is selected, and a line is drawn between it and the wheelchair itself to generate a reverse movement command.

[0115] The final wheelchair control command output module is used to comprehensively analyze the human brain control commands and the original wheelchair commands to generate the final wheelchair control commands, and generate control signals through the PID-sliding mode control algorithm to output to the wheelchair execution terminal.

[0116] The wheelchair actuator is used to receive control signals generated by the final wheelchair command output module and drive the wheelchair to perform corresponding movements.

[0117] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A brain-controlled wheelchair method based on gradient paradigm and human-machine closed loop, characterized in that, include: Based on the spatial distribution characteristics of steady-state visual evoked potentials, a gradient-controlled brain feature evoked system was constructed to obtain characteristic potentials. A brainwave decoding system is constructed using a composite neural network to output brainwave control commands; wherein, the composite neural network is constructed and trained by collecting brainwave data and the characteristic potentials; Based on the EEG control instructions and the original wheelchair control instructions, an initial wheelchair control instruction is obtained, and a final wheelchair control instruction is generated through the PID-sliding mode control method. The wheelchair is then driven to perform corresponding movements through the final wheelchair control instruction. The composite neural network includes: Feature extraction sub-network: used for feature extraction; EEG command output subnetwork: used to output EEG commands; The feature extraction subnetwork includes a three-layer DenseNet network structure, and the EEG command output subnetwork includes a three-layer fully connected network structure. Training the feature extraction subnetwork includes: Target points are evenly distributed in the target area, and users are asked to focus on each target point while their EEG data is collected. The EEG data is then preprocessed. The ratio of the distance between the target point and each stimulus block is measured, and the ratio is normalized to the range of [0,1] to obtain the ideal EEG feature vector of the EEG signal induced by the target location, which is used for subsequent EEG command decoding. Based on the evoked patterns of SSVEP in different visual positions of humans, a feature extraction sub-network structure that can correlate user EEG data and EEG feature vectors is constructed through supervised training. Training the EEG command output subnetwork includes: Establish a relationship function between the target location coordinates and the output brain control command, and obtain the ideal EEG feature vector and its corresponding output brain control command data by uniform sampling method; The EEG command output subnetwork is trained using supervised training based on the ideal EEG feature vector and the corresponding brain control command data. Constructing the EEG decoding system includes: Extracting characteristic potentials and locating the gaze target point; The extraction of feature potentials includes: extracting the evoked intensities of several evoked potentials at different frequencies through the feature extraction subnetwork in the composite neural network to form a feature vector; The process of locating the fixation target point includes: finding the maximum value among the components of the feature vector to determine the user's intention; removing the feature vector components that the user has no intention to control, and using the ratio of component values ​​to correspond to the ratio of distances between the annotation target point and each stimulus pattern to determine the location of the target point being gazed at; obtaining the location coordinates of the target point through calculation, and outputting the EEG control command. The user's intention is to control the movement of the wheelchair or adjust the level of mind control permissions.

2. The brain-controlled wheelchair method based on gradient paradigm and human-machine closed loop as described in claim 1, characterized in that, Constructing the gradient-controlled brain feature induction system includes: By providing visual stimuli to the user through stimulus sources, the location of the user's gaze target point is obtained. The stimulus sources are stimulus patterns used to control the movement and rotation of the wheelchair and patterns used to control the proportion of brain control permissions. Based on the area enclosed by the stimulus pattern used to control the movement and rotation of the wheelchair, i.e. the target area, a user gaze target point is obtained. A Cartesian two-dimensional coordinate system is set at the position of the gaze target point. The position of the gaze target point is described by the two-dimensional coordinates, and the two components of the two-dimensional coordinates are converted into the wheelchair's movement speed and rotation angular velocity, respectively. The stimulus pattern used to control the movement and rotation of the wheelchair changes in the form of a simple harmonic wave. Based on the pattern connecting the line segments used to control the proportion of brain control permissions, the second user gaze target point is obtained, which is used to describe the weight of human brain commands.

3. The brain-controlled wheelchair method based on gradient paradigm and human-machine closed loop as described in claim 1, characterized in that, The initial wheelchair control command is: Among them, v target For the target speed that the wheelchair needs to achieve, v old v is the current speed of the wheelchair, λ is the brain control access coefficient, and v brain v is the speed generated by the current brain-controlled command. esc This is the reverse velocity generated when the wheelchair approaches the obstacle.

4. The brain-controlled wheelchair method based on gradient paradigm and human-machine closed loop as described in claim 1, characterized in that, The PID-sliding mode control method is as follows: Where u(t) represents the control command to be applied at the current moment, including angular velocity and linear velocity; e(t) is the difference between the ideal value and the actual value at the current moment; e(∙) is the error function; K p K i K d These are proportional, integral, and differential gains, respectively.

5. A brain-controlled wheelchair system based on gradient paradigm and human-machine closed loop, used to implement the brain-controlled wheelchair method based on gradient paradigm and human-machine closed loop as described in any one of claims 1-4, characterized in that, include: The system includes a human brain feature induction module, a data acquisition module, a data processing module, a brain feature recognition module, and a human-machine closed-loop control module. The human brain feature induction module is used to construct a brain feature induction system based on stimulus sources; The data acquisition module is used to collect the user's electroencephalogram (EEG) data; The data processing module is used to preprocess the EEG data and construct and train a composite neural network; The brain feature recognition module is used to construct an EEG decoding and classification system based on the composite neural network; The human-machine closed-loop control module is used to construct a human-machine closed-loop control system based on the EEG control commands and the original wheelchair control commands, using the PID-sliding mode control method.

6. The brain-controlled wheelchair system based on gradient paradigm and human-machine closed loop according to claim 5, characterized in that, The human-machine closed-loop control system includes: The brain control command sending module is used to send brainwave control commands to the brain control command receiving module; The human brain control command receiving module is used to receive transmitted brainwave control commands. Wheelchair original control command output module, used to output original wheelchair commands; The final wheelchair control command output module is used to comprehensively analyze the EEG control command and the original wheelchair command to generate the final wheelchair control command, and generate a control signal through the PID-sliding mode control algorithm to output to the wheelchair execution terminal. The wheelchair actuator is used to receive the control signals and drive the wheelchair to perform corresponding movements.

Images