The Research On Human-Computer Interaction Technology Based On Bioelectricity Control

EEG is a non-linear,random,microvolt bioelectrical signal generated by neurons inside the brain and is a comprehensive reflection of the bioelectrical activity of the neuronal cell population on the surface of the scalp.The surface EMG signal of the human is excited by the contraction of superficial neuromuscular cells,causing changes in the surface potential of an electrical signal.At present,there are more and more researches on the brain-computer interaction technology in the world,and the feasibility of the technology is proved in theory and practice.At the same time,due to the certainty and high recognition rate of EMG signal,it has become an ideal source of human-computer interaction control.There are many mature applications of human-computer interaction technology based on the single control of EMG signal.However,the research of EMG and EEG complementary advantage as a synergistic control signal source to control external devices and equipment is relatively few.As the surface EMG signal can directly reflect the body's limb activity,it is very suitable for anthropomorphic control;The brain as the nerve center of a man,EEG signals are collected through the cerebral cortex contains a wealth of control information,and can be used as a controlled signal source for the controlled equipment.Based on the research status of EEG and EMG,this paper makes a detailed study and analysis of signal collection,preprocessing,feature extraction and pattern recognition.At the same time,a multi mode teleoperation wheeled car based on EEG and EMG control is developed,which provides an online verification platform for EEG and EMG analysis algorithms presented in this paper.In this paper,the EEG signals are used to control the four kinds of moving state of wheeled car,that is,turn left,turn right,advance and retreat;Six-axis attitude sensor controls six-degree-of-freedom mechanical arm on wheeled car;The EMG signals complete the grip of the claw.The main contents and related study results of this paper are induced as follows:(1)In view of the traditional blind source separation,it is not possible to remove all artifacts from the prior information of the original signal,and a method of removing the artifacts of EEG signals based on the tangent function denoise source separation is proposed.Based on the prior information of the observed signal,such as sparse time domain,sparse frequency domain and sequential structure,the method can be used to construct a specific noise reduction function and separate the component signals from the observed signals.In this paper,we selected a piece of original signals including EOG artifacts and EEG signal as off-line analysis objects.By using the denoise source separation algorithm,the effect of removing artifacts were better than those using Fast ICA algorithm.According to the characteristics of EMG,the Denoising method based on wavelet transform is chosen to remove noise.(2)The wavelet transform is used to extract alpha and beta rhythm waves from EEG signals,and the corresponding average normalized energy values are obtained as characteristic 1;According to the nonlinear characteristics of EEG signals,a fast approximate entropy feature extraction method based on binary distance matrix is proposed as feature 2;The two features are then combined to form the combined eigenvector of the EEG.According to the four aspects of time domain,frequency domain,time-frequency domain and entropy of EMG,four characteristics of WAMP,MF,EWT and PE are selected.The eigenvectors of EMG signals are obtained after the combination of the four features.In this paper,we propose a kind of evaluation index based on the between class scatter matrix and the within class scatter matrix,and the advantages and disadvantages of all kinds of characteristics are evaluated.(3)In view of the rejection phenomenon that traditional one-versus-rest(OVR)classifiers may appear,this paper proposes a support vector machine classification method based on directed acyclic graph(DAG-SVM),which is used to identify four types of motorized imaging EEG.Tree-structured classifier can effectively avoid the phenomenon of rejection;Using SVM to classify and identify the EMG signals of two kinds of action modes of hand extension and grasp.The off-line analysis showed that the average classification accuracy of the combined features of EEG signals was 67.19%,which was better than that of single features.The average classification accuracy of combined features of EMG signals was 99%,which was better than that of single feature classification rate.(4)A remote operation wheeled with EEG and EMG control is designed and implemented.Based on the VS2010 platform,the PC client software is developed,and the preprocessing,feature extraction and recognition classification of EEG/EMG signal are completed by using the client to invoke MATLAB interface.And the classification results into instructions,via Wi Fi remote transmission to the wheeled car,control the movement of the car and manipulator claw operation.The result of the recognition of EEG signals with left hand,right hand,feet and tongue is to control the car turn left,turn right,advance and retreat;Six-axis attitude sensor data control the movement of six-degree-of-freedom manipulator,and control the hand-claw opening and closing by the recognition result of the hand extension and grasp,and carry on the on-line control experiment.