Audio-visual Guidance Of Motor Imagery EEG Signals Identification Research

Neural electrophysiological phenomenon in the cerebral cortex in the time of motor imagery(MI)is similar to the actual movement,it can directly activate the sensorimotor cortex and help the sensorimotor cortex to restructure or reconstruct.So MI can be used to help patients with motor dysfunction caused by brain diseases to recover.And the recognition of EEG signals of different action images can also be used as a control strategy to realize the Brain-Computer interaction.Therefore,the analysis of motor cortical state and the pattern recognition of different action images both have realistic significance and scientific research value.In this paper,author researches and analyzes EEG signals in different visual and auditory guidance conditions in the current study of motor cortical state in the recognition and movement of EEG,and mainly focuses on the state analysis of motor cortex and the pattern recognition of EEG signal of different action images.The main research content and innovation points include the following aspects:(1)The design of four groups different visual and auditory guidance conditions MI experiments.The synchronously playing action video is selected as the visual guidance,the voice based on auditory cognition that sounds like "left" and "right" is selected as the auditory guidance.The motor cortical state and action pattern recognition of different guidance conditions are compared and analyzed.(2)The analysis of motor cortical state.Firstly,the AR power spectrum model is selected to analyze the change of signal energy in the whole frequency band,and to observe the performance of the motor cortical state in the whole frequency domain.In terms of the mu rhythm signal is generated by the motor cortex and the neural mechanism of ERD/ERS phenomenon,the time domain and frequency domain ERD quantization values of the mu rhythm signal is selected as active measure of motor cortex to analyze the state of motor cortex in different visual and auditory guidance condition.The results show that the activity level of motor cortex was highest in the audio-visual guidance,the visual guidance was secondary,the auditory guidance was low,and the single MI was the lowest,and the effect of vision on cortical state is greater than that of auditory.(3)Feature extraction of EEG signal.Firstly,a CSP method based on SWPT is proposed to solve the problem of low classification accuracy and the lack of frequency domain information in the traditional CSP,The EEG signal is decomposed and reconstructed by SWPT,and the four frequency bands signals in each channel are selected,and the three channel have 12 signals as input signals of the CSP method,the classification results show that SWPT-CSP is better than CSP.Secondly,for multi-channel EEG signal feature lead to large amount of calculation and features lack robustness in computing MMSE,using an improved MMSE algorithm,simulation results show that the IMMSE feature of EEG can be well distinguished.(4)Design classifier.In order to solve the problem of low accuracy in the pattern classification of EEG at present,the differential evolution algorithm is introduced to optimize LS-SVM,The recognition rate is greatly improved in two features.When the feature is IMMSE,compared with SVM and LS-SVM,the recognition rate is improved 2.75% and 4.25%;when the feature is SWPT-CSP,the recognition rate is improved 4.75% and 5.25%.The classification recognition rate of different features and different guidance conditions is analyzed,the result show that among the three features of CSP,SWPT-CSP and IMMSE,classification accuracy of SWPT-CSP is the best,IMMSE secondary,CSP lowest.In four different guidance conditions,classification recognition rate: audio-visual guidance > vision guided > auditory guidance > single MI,and the effect of vision on classification recognition rate is greater than that of auditory.