Design And Research Of Brain Controlled Wheelchair Based On Steady State Visual Evoked Potential

Human diseases or major accidents can lead to the loss of some motor functions,which greatly limits the patient’s ability to move and take care of themselves in daily life.The emergence of brain-controlled wheelchairs can effectively solve this problem,but braincontrolled wheelchairs have the problems of slow motion consistency between brain-computer interface systems and wheelchairs,and low recognition rate of EEG signals.In view of the above problems,this paper designs a brain-controlled wheelchair based on steady-state visual evoked potential,and realizes the communication between the patient’s brain and the wheelchair through the brain-computer interface system.This paper mainly designs the visual stimulation system of brain-computer interface system,studies the feature extraction and feature recognition algorithm based on visual steady-state evoked potential(SSVEP)signal,and realizes the control of wheelchair movement by EEG signal by building brain-controlled wheelchair system.This paper first introduces the components of the brain-computer interface system and the principle of Steady State Visual Evoked Potentials signal.The visual stimulation software based on SSVEP is programmed and four stimulation frequencies are selected.Then Matlab is used to remove the artifacts of the SSVEP signal,and the Butterworth filter is designed to perform power frequency filtering on the SSVEP signal after the artifacts are removed to complete the preprocessing of the SSVEP signal.In order to improve the signal recognition rate of brain-controlled wheelchair instructions,wavelet transform-canonical correlation analysis algorithm is introduced to process SSVEP signals.Four algorithms,canonical correlation analysis,multi-set canonical correlation analysis,filter bank canonical correlation analysis and continuous wavelet transform-canonical correlation analysis,are used to conduct offline experiments on the data set.The experimental results show that the recognition rate of continuous wavelet transform-canonical correlation analysis SSVEP signal is better than other canonical correlation analysis algorithms.When the time window is 2 seconds,the recognition rate of continuous wavelet transform-canonical correlation analysis SSVEP signal reaches 88%,which effectively improves the recognition accuracy of SSVEP signal.Aiming at the problem of low efficiency of canonical correlation analysis algorithm in processing SSVEP signal short time window,a convolutional neural network is established to process SSVEP model signal.The existing data set is expanded by data enhancement method and the neural network is trained by gradient descent method.The optimal hyperparameters for processing SSVEP signals are obtained and the convolutional neural network model is constructed to conduct offline experiments on the data set.The experimental results show that the convolutional neural network has a significant effect on processing SSVEP signals in a short time window.When the time window is 1 second,the signal recognition rate of the convolutional neural network reaches 76.3 %,which effectively improves the recognition accuracy of SSVEP signals in a short time window.Finally,in order to verify the effectiveness of the SSVEP brain-computer interface system and algorithm,this paper builds a brain-controlled wheelchair prototype and designs a braincontrolled wheelchair control system.The online experiment of brain-controlled command accuracy is carried out,which proves the effectiveness of the brain-controlled wheelchair and the feasibility of the control scheme,and the movement of the brain-controlled wheelchair is preliminarily realized.