阅读说明：本技术 一种基于运动想象脑电与头姿的智能轮椅及控制方法 (Intelligent wheelchair based on motor imagery electroencephalogram and head posture and control method ) 是由 王强 徐国政 孙星 高翔 谭彩铭 朱博 于 2021-08-23 设计创作，主要内容包括：本发明公开了一种基于运动想象脑电与头姿的智能轮椅及控制方法,智能轮椅装置包括电动轮椅和附加装置脑电采集设备、头姿采集设备、颈部肌电采集设备、头部姿态估计模块、虚拟光标控制模块、轮椅人机交互接口模块、疲劳感知模块和智能轮椅。控制方法包括：(1)运动想象脑电控制模式使用感觉运动节律控制虚拟光标,头姿交互控制模式使用头部姿态控制虚拟光标；(2)根据用户颈部肌肉的疲劳状态切换运动想象脑电与头姿两种控制方式；(3)根据虚拟光标的坐标位置并通过差分运动学模型计算轮椅左右轮对应的转速。利用该控制方法及装置,可连续调节智能轮椅的速度和转向角度,缓解了用户操作时颈部肌肉的疲劳程度,提升了智能轮椅的连续性和舒适性。(The invention discloses an intelligent wheelchair based on motor imagery electroencephalogram and head posture and a control method. The control method comprises the following steps: (1) the motor imagery brain electric control mode uses the sensory motor rhythm to control the virtual cursor, and the head posture interaction control mode uses the head posture to control the virtual cursor; (2) switching two control modes of motor imagery electroencephalogram and head posture according to the fatigue state of the neck muscles of the user; (3) and calculating the corresponding rotating speeds of the left wheel and the right wheel of the wheelchair through a differential kinematics model according to the coordinate position of the virtual cursor. By utilizing the control method and the control device, the speed and the steering angle of the intelligent wheelchair can be continuously adjusted, the fatigue degree of neck muscles during operation of a user is relieved, and the continuity and the comfort of the intelligent wheelchair are improved.)

1. The utility model provides an intelligence wheelchair based on motor imagery brain electricity and head appearance, includes electronic wheelchair main part, its characterized in that still includes the additional structure of installing in the main part: the system comprises an electroencephalogram acquisition device, a head posture acquisition device, a neck myoelectricity acquisition device, a head posture estimation module, a virtual cursor control module, a wheelchair human-computer interaction interface module and a fatigue sensing module; the Kinect depth camera, the PC controller tray, the storage battery and the inverter;

the electroencephalogram acquisition equipment is installed on the back of the electric wheelchair and is powered up through a storage battery and an inverter; the Kinect depth camera is installed right in front of a user, electroencephalogram and head posture signal processing and wheelchair control are processed by the PC controller, and the PC controller is placed on a tray right in front of the user, so that visual feedback and man-machine interaction are facilitated;

the electroencephalogram signals collected by the electroencephalogram collecting equipment are transmitted to the virtual cursor control module through the USB interface, then are transmitted to the wheelchair man-machine interaction interface through TCP communication, and finally, control instructions are transmitted to the electric wheelchair through a serial port communication mode;

the head posture acquired by the head posture acquisition equipment is transmitted to a head posture estimation module through a USB interface, is transmitted to a wheelchair man-machine interaction interface through TCP communication, and finally transmits a control instruction to the electric wheelchair through a serial port communication mode;

neck electromyographic signals collected by the electromyographic collecting device are transmitted to the fatigue sensing module through TCP communication, and electroencephalogram control and head posture control are judged and switched according to a fatigue sensing result.

2. The method for controlling the intelligent wheelchair as claimed in claim 1, wherein on one hand, the head pose depth image captured by the Kinect sensor is further mapped to the human-computer interaction interface of the robot wheelchair to generate a control instruction after the head pose real-time pose is estimated by the head pose estimation module, so as to realize the head pose interaction control of the wheelchair; on the other hand, a Neuroscan brain electric controller collects motor imagery brain electric signals in real time, the motor imagery brain electric signals are used for controlling a virtual cursor after being processed in a pre-stage mode, the controlled virtual cursor is synchronously mapped to a human-computer interaction interface of the robot wheelchair, and continuous brain electric control instructions are generated to achieve interaction of the wheelchair; after neck surface electromyographic signals collected by the Delsys electromyographic sensor are evaluated by the fatigue sensing module, whether a switching instruction is generated or not is determined according to the fatigue state, and switching between two interactive modes of head posture and electroencephalogram is realized; the method specifically comprises the following steps:

s1, under the head posture control mode, acquiring the head posture of the user in real time by using the Kinect depth camera, and the method specifically comprises the following steps:

s11, estimating the head posture of the subject by using a random forest fusion closest point iterative algorithm;

s12, mapping the real-time head posture coordinates of the subject to a wheelchair control module;

s13, continuously outputting a speed instruction of the wheelchair in real time according to the change of the head posture coordinate;

s2, under the motor imagery electroencephalogram control mode, the motor imagery electroencephalogram control needs to obtain the sensorimotor rhythm parameters of the testees through a screening experiment, and the experimental data is analyzed to find the respective optimal parameter configuration of each tester;

s3, converting the electroencephalogram signals of the subject into control instructions of a virtual cursor through feature extraction and feature conversion;

s4, synchronously mapping the virtual cursor controlled in the step S3 to a human-computer interaction interface of the robot wheelchair, and generating speed instructions of left and right wheels of the wheelchair to drive the wheelchair to move;

s5, monitoring the fatigue physiological response of the neck muscles of the user in real time, judging the fatigue state of the neck muscles of the subject according to the fatigue evaluation model, and determining whether to switch between the motor imagery electroencephalogram and head posture interactive control modes.

3. The control method of claim 2, wherein in step S2, the electroencephalogram acquisition device is a SynAmp2 amplifier and a 64-channel Quik-Cap electrode developed by Neuroscan, of which 18 channels, FC6, FC4, FC2, FC5, FC3, FC1, C6, C4, C2, C5, C3, C1, CP6, CP4, CP2, CP5, CP3 and CP1 are selected, and the Scan4.5 acquisition software is used to acquire the electroencephalogram signals of the user.

4. The control method according to claim 2, wherein in step S3, the following is specifically performed: in the step, a common average reference filter is used for carrying out spatial filtering on the acquired electroencephalogram signals, the average value of all electrodes except the central electrode is calculated and used as a reference, and the formula is

Wherein s is_h(t) is the original potential of the target electrode at time t,for potentials filtered in CAR space, s_i(t) is the potential at the ith electrode recorded;

then, time filtering is carried out by utilizing a maximum entropy method, namely a burg algorithm, based on an autoregressive model, and time domain signals of the Mu rhythm and the Beta rhythm are converted into frequency domain characteristics for analysis; the AR model is represented as:

wherein u (n) is zero as a mean and δ as a variance²P is the order of a_kAre AR model parameters. After calculating the parameters of the AR model, calculating the power spectrum:

according to the change of spectral energy between two states of comparing motor imagery and rest of a user, finding out the channel and frequency representation with the maximum difference which are most related to the motor imagery of the user, and taking the electroencephalogram signal amplitude at the position as the final signal characteristic;

and finally, converting the signal characteristics into a control signal of the cursor by a linear regression method, wherein the vertical motion component formula of the virtual cursor is as follows:

M_v＝ω_rvR_v+ω_lvL_v+b_v

wherein R is_vAmplitude of a particular frequency band on the right side of the brain, L_vThe amplitude of the left specific frequency band; omega_rvAnd ω_rvIs the corresponding weight coefficient; the formula of the horizontal motion component of the virtual cursor is as follows:

M_h＝ω_rhR_h+ω_lhL_h+b_h

wherein R is_hAmplitude of a particular frequency band on the right side of the brain, L_hThe amplitude of the left specific frequency band; omega_rhAnd ω_lhIs the corresponding weight coefficient; coefficient ω in the initial state_rv＝ω_rv＝1，ω_rh＝1，ω_lh-1; after one experienceAnd after the whole experiment process, updating the weight coefficient and the offset by adopting a minimum mean square error algorithm according to the previous experiment data.

5. The control method according to claim 2, wherein in step S4, the speed command of the wheelchair is continuously outputted in real time according to the change of the virtual cursor position coordinates (x, y) by the following formula:

wherein x and y are coordinates of the virtual cursor, x 'and y' are coordinates of the mapped control cursor, v_lAnd v_rThe moving speed of the axle center of the left wheel and the right wheel, R is the radius of the driving wheel, L is the wheel track of the driving wheel, and v and omega are the linear speed and the angular speed of the wheelchair; and decomposing the generated speed instruction according to the differential kinematics model and sending the speed instruction to the PID controller at the bottom layer to adjust the rotating speed of the left wheel motor and the right wheel motor and drive the wheelchair to move.

6. The control method according to claim 2, wherein in the step S5, the fatigue evaluation model determines that the user is in a fatigue state if the integral myoelectric value is increased and the average power frequency is decreased during an evaluation period according to the integral myoelectric value and the average power frequency of the physiological response of the fatigue of the neck muscle, and otherwise, the user is in a non-fatigue state; when the user is in a fatigue state, switching from a head posture interaction mode to a motor imagery electroencephalogram control mode; when the user is in a non-fatigue state, the motor imagery electroencephalogram control mode is switched to the head posture interaction mode.

Technical Field

The invention relates to the technical field of brain-computer interfaces and machine vision, in particular to a method and a system for continuously controlling an intelligent wheelchair based on brain and head posture interaction by motor imagery.

Background

With the improvement of the attention of the current society to the elderly and disabled, how to meet the travel demands of the above-mentioned groups becomes a social hotspot problem to be solved urgently. In recent years, intelligent wheelchairs play an important role in solving the travel problems of the elderly and disabled. The brain-computer interface is an emerging technology, and one of the final research goals of the brain-computer interface is to realize control over external equipment by decoding brain activities of users in real time, so that the brain-computer interface can help the high-level paraplegic and hemiplegic users to establish contact with the outside world again. Motor imagery brain electrical signals are typical spontaneous brain electrical signals, and when a subject imagines a certain specific motion without making an actual limb motion, a sensory motor area of cerebral cortex can generate corresponding potential changes. Phenomena of increasing and decreasing variation according to the corresponding signal are called event-related synchronization and event-related desynchronization phenomena. Performing different motor imagery actions will produce corresponding changes in the Mu and Beta rhythms at different locations. The event-related synchronization and event-related desynchronization phenomena are used as important bases for judging different motor imagery actions, so that the subjective will of a subject can be more intuitively reflected, and a motor imagery brain-computer interface is widely used. The head gesture interaction utilizes the camera to acquire the specific motion of the head of the user in real time and converts the specific motion into a control instruction, so that the control of external equipment is realized. The head posture interactive control can help a user to complete the human-computer interaction of the wheelchair through intuitive head movement, but secondary damage is easily caused to the user due to the fatigue of neck muscles. The traditional brain-computer interface control intelligent wheelchair is mainly based on the control of four-classification identification of electroencephalogram signals. Usually, only one section of electroencephalogram signal can be collected, and after the electroencephalogram signals are classified, a discrete forward, backward, left-turn or right-turn control instruction is obtained, so that the difference between the riding experience and the control of other traditional modes is large, the speed cannot be adjusted, and the continuity and the comfort are not realized.

The continuous control method provided by the invention is based on the motor imagery electroencephalogram and head posture interaction, and monitors the fatigue state of neck muscles of a subject in real time as a condition for switching to motor imagery electroencephalogram control. In the head posture control mode, when fatigue of the neck muscles of the user is detected, the control mode is switched to the motor imagery control mode. Under the motor imagery control mode, firstly, the decoded motor imagery electroencephalogram signals are converted into control instructions of a virtual cursor, then the cursor is mapped to a wheelchair control interface in real time, and the rotating speeds of the left wheel and the right wheel of the intelligent wheelchair are continuously output in real time according to the position of the cursor.

Disclosure of Invention

The purpose of the invention is as follows: the invention aims to provide an intelligent wheelchair based on motor imagery electroencephalogram and head posture and a control method, which are used for monitoring the fatigue physiological response of neck muscles of a user in real time and switching two control modes under the condition of a fatigue evaluation result. The motor imagery electroencephalogram control mode decodes electroencephalogram signals of a user during motor imagery and converts the electroencephalogram signals into a control instruction of a virtual cursor, the cursor is mapped to a wheelchair control interface in real time, and the control instruction is continuously output to a wheelchair in real time according to the position of the cursor, so that the wheelchair has the capability of adjusting speed and steering angle.

The technical scheme is as follows: the invention relates to an intelligent wheelchair based on motor imagery electroencephalogram and head posture and a control method, wherein a wheelchair system comprises: the wheelchair comprises brain electricity acquisition equipment, head posture acquisition equipment, neck myoelectricity acquisition equipment, a head posture estimation module, a virtual cursor control module, a wheelchair man-machine interaction interface module, a fatigue sensing module and an intelligent wheelchair device. On one hand, after a head posture depth image captured by the Kinect sensor is subjected to head posture estimation by a head posture estimation module, the head posture depth image is further mapped to a human-computer interaction interface of the robot wheelchair to generate a control instruction, so that head posture interaction control of the wheelchair is realized; on the other hand, a Neuroscan brain electric controller collects motor imagery brain electric signals in real time, the motor imagery brain electric signals are used for controlling a virtual cursor after being processed in a pre-stage mode, the controlled virtual cursor is synchronously mapped to a human-computer interaction interface of the robot wheelchair, and continuous brain electric control instructions are generated to achieve interaction of the wheelchair; after neck surface electromyographic signals collected by the Delsys electromyographic sensor are evaluated by the fatigue sensing module, whether a switching instruction is generated or not is determined according to the fatigue state, and switching between two interactive modes of head posture and electroencephalogram is achieved.

The method mainly comprises the following steps:

s1, acquiring head image information of the user in real time by using the Kinect depth camera in a head posture control mode;

(1) estimating the head posture of the subject by using a random forest fusion closest point iterative algorithm;

(2) mapping real-time head pose coordinates of the subject to a wheelchair control module;

(3) and continuously outputting the speed instruction of the wheelchair in real time according to the change of the head posture coordinate.

S2, obtaining sensory-motor rhythm parameters of the testees through screening experiments in motor imagery electroencephalogram control, and analyzing experimental data to find out the optimal parameter configuration of each tester;

in the step, the electroencephalogram acquisition equipment is a Synamp2 amplifier and a 64-channel Quik-Cap electrode developed by Neuroscan, wherein FC6, FP4, FC2, FC1, FC3, FC5, C6, C4, C2, C5, C3, C1, CP6, CP4, CP2, CP5, CP3 and CP1 channels are selected, and Scan4.5 acquisition software is used for acquiring electroencephalogram signals of a user.

S3, converting the electroencephalogram signals of the subject into control instructions of a virtual cursor through feature extraction and feature conversion;

in the step, a common average reference filter is used for carrying out spatial filtering on the acquired electroencephalogram signals, the average value of all electrodes except the central electrode is calculated and used as a reference, and the formula is

Wherein s is_h(t) is the original potential of the target electrode at time t,for potentials filtered in CAR space, s_i(t) is the potential at the ith electrode recorded.

And then, performing time filtering by using a maximum entropy method (burg algorithm) based on an autoregressive model, and converting time domain signals of the Mu rhythm and the Beta rhythm into frequency domain characteristics for analysis. The AR model may be represented as:

wherein u (n) is zero as a mean and δ as a variance²P is the order of a_kAre AR model parameters. After calculating the parameters of the AR model, calculating the power spectrum:

according to the change of spectral energy between two states of comparing the motor imagery and rest of the user, the channel and frequency representation with the maximum difference are found to be most relevant to the motor imagery of the user, and the amplitude of the electroencephalogram signal at the position is used as the final signal characteristic.

And finally, converting the signal characteristics into a control signal of the cursor by a linear regression method, wherein the vertical motion component formula of the virtual cursor is as follows:

M_v＝ω_rvR_v+ω_lvL_v+b_v

wherein R is_vAmplitude of a particular frequency band on the right side of the brain, L_vThe amplitude of the left specific band. Omega_rvAnd ω_rvAre corresponding weight coefficients. The formula of the horizontal motion component of the virtual cursor is as follows:

M_h＝ω_rhR_h+ω_lhL_h+b_h

wherein R is_hAmplitude of a particular frequency band on the right side of the brain, L_hThe amplitude of the left specific band. Omega_rhAnd ω_lhIs the corresponding weightAnd (4) the coefficient. Coefficient ω in the initial state_rv＝ω_rv＝1，ω_rh＝1，ω_lhIs-1. After a complete experimental process, updating the weight coefficient and the offset by adopting a minimum mean square error algorithm according to previous experimental data.

And S4, synchronously mapping the virtual cursor controlled in the step S3 to a human-computer interaction interface of the robot wheelchair, and generating speed commands of left and right wheels of the wheelchair to drive the wheelchair to move.

In this step, the speed command of the wheelchair is continuously output in real time according to the change of the virtual cursor position coordinates (x, y). The formula is as follows:

wherein x and y are coordinates of the virtual cursor, x 'and y' are coordinates of the mapped control cursor, v_lAnd v_rThe moving speed of the axle center of the left wheel and the right wheel, R is the radius of the driving wheel, L is the wheel track of the driving wheel, and v and omega are the linear speed and the angular speed of the wheelchair. And decomposing the generated speed instruction according to the differential kinematics model and sending the speed instruction to the PID controller at the bottom layer to adjust the rotating speed of the left wheel motor and the right wheel motor and drive the wheelchair to move.

S5, monitoring the fatigue physiological response of the neck muscles of the user in real time, judging the fatigue state of the neck muscles of the subject according to the fatigue evaluation model, and determining whether to switch between the motor imagery electroencephalogram and head posture interactive control modes.

In the step, the fatigue evaluation model mainly judges that the user is in a fatigue state if the integral myoelectric value is increased and the average power frequency is reduced in an evaluation period according to the integral myoelectric value and the average power frequency of the physiological response of the fatigue of the neck muscle, and otherwise, the user is in a non-fatigue state. When the user is in a fatigue state, switching from a head posture interaction mode to a motor imagery electroencephalogram control mode; when the user is in a non-fatigue state, the motor imagery electroencephalogram control mode is switched to the head posture interaction mode.

Drawings

FIG. 1 is a block diagram of an intelligent wheelchair;

FIG. 2 is a structural diagram of a continuous control system of an intelligent wheelchair based on interaction of electroencephalogram and head posture through motor imagery;

FIG. 3 is a schematic view of a wheelchair control module interface;

fig. 4 is a flow chart of a motor imagery electroencephalogram and head posture interactive control switching method.

Advantageous effects

The invention provides an intelligent wheelchair continuous control method and system based on motor imagery electroencephalogram and head posture interaction, wherein the electroencephalogram control intention of a user is converted into a control instruction of a virtual cursor, a controlled cursor is mapped to a wheelchair man-machine interaction interface in real time, and the control instruction is continuously output to a wheelchair in real time according to the position of the cursor; meanwhile, according to the fatigue state of the neck muscles of the user monitored in real time, switching is carried out in two interactive control modes of motor imagery electroencephalogram and head posture, the muscle fatigue of the user is relieved, and the continuity and the comfort of the user in driving the intelligent wheelchair are improved.

Detailed Description

In order to clearly illustrate the technical contents of the present invention, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments, wherein the specific embodiments and the description are only used for explaining the present invention, but not for limiting the present invention.

Referring to fig. 1, the structure of the intelligent wheelchair used in the present invention is a structure of the intelligent wheelchair having a basic electric wheelchair structure as a subject and an auxiliary structure installed on a main body, including a Kinect depth camera 101, a PC controller tray 102, a wheelchair main body 103, front wheels 104, a motor 105, a syncamps 2 brain electrical acquisition device 106, a storage battery and inverter 107, and rear wheels 108. As shown in the figure, SynAmps2 brain electricity collection equipment is installed on the back of the intelligent wheelchair and is powered by a storage battery through an inverter. The Kinect depth camera is mounted right in front of the user, ensuring that the entire head of the user is contained in the captured image. The brain electricity and head posture signal processing and the wheelchair control are processed by a PC controller, and the PC controller is placed on a tray right in front of a user, so that visual feedback and man-machine interaction are conveniently carried out.

Referring to fig. 2, a structure diagram of an intelligent wheelchair continuous control system based on combination of motor imagery electroencephalogram and head posture is shown, and the main implementation scheme is as follows: after the head posture depth image captured by the Kinect sensor is used for estimating the real-time head posture through the head posture estimation module, the head posture depth image is further mapped to a human-computer interaction interface of the robot wheelchair to generate a control instruction, and head posture interaction control of the wheelchair is achieved; a Neuroscan brain electrical controller collects motor imagery brain electrical signals in real time, the motor imagery brain electrical signals are used for controlling a virtual cursor after being processed in a pre-stage mode, the controlled virtual cursor is synchronously mapped to a human-computer interaction interface of the robot wheelchair, and continuous brain electrical control instructions are generated to realize the interaction of the wheelchair; after neck surface electromyographic signals collected by the Delsys electromyographic sensor are evaluated by the fatigue sensing module, whether a switching instruction is generated or not is determined according to the fatigue state, and switching between two interactive modes of head posture and electroencephalogram is realized;

referring to fig. 3, which is a schematic view of an interface of a wheelchair human-computer interaction interface control module, a rectangular area is a motion area of a virtual cursor, and a circular area is a cursor motion area mapped by the wheelchair control module.

Referring to fig. 4, a flow chart of a motor imagery electroencephalogram control and head posture interactive control switching method is shown, and the fatigue state of neck muscles of a subject is monitored in real time; when the fatigue state is judged, the head posture interactive control mode is switched to motor imagery electroencephalogram continuous control; and when the judgment result is that the muscles are in the recovery state, switching to head posture interactive control.

The following describes an implementation of an embodiment of an intelligent wheelchair continuous control method based on motor imagery electroencephalogram and head posture more specifically with reference to the accompanying drawings.

S1, the testee wears a 64-lead Quik-Cap connected with a SynAmps2 amplifier and an electromyographic sensor in a Trigno wireless surface electromyographic test system, conductive paste is injected into FC6, FP4, FC2, FC1, FC3, FC5, C6, C4, C2, C5, C3, C1, CP6, CP4, CP2, CP5, CP3 and CP1 channels of the electrode Cap, the electrical brain signals of the user are collected by Scan4.5 collection software, and a Kinect depth camera is placed right in front of the user.

S2, firstly, starting a head posture interaction control mode;

s21, estimating the head pose of the subject by using a random forest fusion closest point iterative algorithm;

s22, mapping the real-time head posture coordinates of the subject to a wheelchair control module;

and S23, continuously outputting the speed command of the wheelchair in real time according to the change of the head posture coordinate. The user controls the wheelchair to move continuously by raising the head, lowering the head, leftwards turning the head and rightwards turning the head.

And S3, converting the electroencephalogram signals of the subject into the control instructions of the virtual cursor through feature extraction and feature conversion.

S31, feature extraction is firstly carried out in BCI2000 electroencephalogram processing software. The user wants to do left-hand or right-hand movement according to the prompt, the common average reference filter is used for carrying out spatial filtering on the acquired electroencephalogram signals, the average value of all electrodes except the central electrode is calculated and used as a reference, and the formula is that

S32, calculating the spectrum of the signal after spatial filtering by using an AR model, where the AR model can be expressed as:after the parameters of the AR model are calculated, the power spectrum is calculated, and the formula is as follows:

s33, analyzing and comparing the frequency spectrum information of the subject in the rest state and the ideal left-hand or right-hand state, finding out the channel and the frequency with the maximum difference, and using the amplitude at the position as the finally extracted signal characteristic to form the virtual cursor horizontal motion component, wherein the formula is as follows:

M_h＝ω_rhR_h+ω_lhL_h+b_h

analyzing and comparing the frequency spectrum information of the subject in a rest state and an imaginary two-hand movement or relaxation state, finding out a channel and a frequency with the maximum difference, and forming a virtual cursor vertical movement component by taking the amplitude at the position as the finally extracted signal characteristic, wherein the formula is as follows:

M_v＝ω_rvR_v+ω_lvL_v+b_v

coefficient ω in the initial state_rv＝ω_rv＝1，ω_rh＝1，ω_lhIs-1. After a complete experimental process, updating the weight coefficient and the offset by adopting a minimum mean square error algorithm according to previous experimental data.

And S4, recording myoelectric data of 6 sensors of the user' S pinctada at intervals of 2 minutes in the head posture interaction control process, and recording myoelectric data of 30 seconds at each time to be used as fatigue degree evaluation.

Specifically, in an evaluation period, a sliding window calculation with an overlap rate of 50% is performed with 1000 data points as a time window and a step size of 500 points. Carrying out two kinds of calculation each time to respectively obtain an integral myoelectric value and an average power frequency within each window time, and sequentially calculating to obtain all integral myoelectric values and average power frequencies within complete 30 seconds; then, performing one-dimensional least square fitting on the data in each evaluation period, and judging the change of the corresponding evaluation index in the evaluation period according to the slope of the fitted straight line; when the integral myoelectricity value of the user rises and the average power frequency is reduced, judging that the user is in a fatigue state, and switching to a motor imagery brain electric control mode; and judging the head posture interactive control mode to be in a non-fatigue state under other conditions, and continuously keeping the head posture interactive control mode.

And S5, transmitting the position of the virtual cursor in the step S3 to the wheelchair control module in real time through TCP communication, and generating speed commands of the left wheel and the right wheel of the wheelchair to drive the wheelchair to move. The speed command of the wheelchair is continuously output in real time according to the change of the virtual cursor position coordinates (x, y), as shown in fig. 3. The formula is as follows:

and decomposing the generated speed instruction according to the differential kinematics model and sending the speed instruction to the PID controller at the bottom layer to adjust the rotating speed of the left wheel motor and the right wheel motor and drive the wheelchair to move.

The above description is only exemplary of the present invention and should not be taken as limiting the invention, as any modification, equivalent replacement, or improvement made within the spirit and scope of the present invention should be included in the present invention.