Research On Algorithm Based On Motor Imagery Brain-computer Interface

The Brain Computer Interface(BCI)is a system that can provide a communication control method between the brain and the outside world without relying on the normal output pathway of the brain peripheral nerves and muscle tissues.In recent years,BCI technology has become a hot research field of biomedical engineering,and has been applied in various fields,which is followed by higher requirements for the accuracy and stability of BCI systems.The BCI based on motor imagery EEG signals is an important research direction of BCI,and it is also the main content of this paper.Based on the full understanding of national and international research and development of BCI,this paper systematically studied a left-right hand motor imagery BCI based on EEG signals.It discussed from offline analysis problems to real-time online problems.The research aim of this paper is to effectively improve the classification accuracy of left and right hand motor imagery.To achieve this goal,adaptive parameterless empirical wavelet transform(APEWT)and convolutional neural network is used to extract features of motor imagery EEG signals.Then due to subjects specificity,there are different optimal time domain and frequency domain features during different subjects.In addition,The classification of left and right hand motor imagery EEG signals is achieved by weighted integrated classification methods for different time periods and frequency bands.The main work of the paper is as follows:1)The mechanism of motor imagery EEG signals was analyzed.According to nonlinearity and non-stationarity of EEG signals,a new decomposition method called adaptive parameterless empirical wavelet transform was used to decompose the EEG signals.And a selective integrated classification model was proposed to boost classification accuracy.2)Considering the characteristics of different subjects with different sensitivity to different features,the convolutional neural network was used to automatically extract features.Then,the different features are used as the input of convolutional neural network.Finally,identification of left and right hand motor imagery can be achieved by integrated classification.3)The online motor imagery BCI classification and recognition system based on EEG signals was designed,which realized the identification of idle state and solved the problem of continuous operation of BCI.Experiments show that after a short training period on the online platform,the subjects can freely switch between different states and had a higher classification accuracy.