Algorithm Analysis Based On Motion Imaginary Brain Computer Interface

The 21 st century is recognized by the world as an era of biological sciences and brain sciences.The multidisciplinary and multi-level comprehensive research on the higher cognitive functions of the human brain has shown an exponential growth trend.Driven by the global brain research program,China has proposed the strategic deployment of "Brain science and Brain-like research" and the "Chinese brain plan",in order to achieve international leading results in the three frontier fields of brain science,early diagnosis and intervention of brain diseases,and brain-like intelligence calculation.The Brain-Computer Interface(BCI)is a system for communication and control between the human brain and a computer or other electronic device that does not depend on the human peripheral nervous system and muscle tissue.The BCI system based on motion imaging mainly uses EEG signals as signal input,uses computer to perform signal processing to judge the type of motion imagination and converts it into control commands,thereby realizing the communication and control functions between the human brain and external devices.However,the ability of the system to quickly and efficiently extract task-related features is currently the biggest issue facing researchers.With advances in signal acquisition techniques,the huge data of EEG signals and existing unrelated features limit the performance of the classifier.In order to solve these problems,this study proposes two efficient signal processing frameworks based on the binary MI-BCI system,namely motor imagery brain-computer interface algorithm processing framework based on feature selection and motor imagery brain-computer interface system based on channel adaptive selection.Stockwell transform and Bayesian linear discriminant analysis are applied to feature extraction and classification,respectively.In the feature selection process,the most relevant classification features are extracted by using Genetic Algorithm(GA),and selecting the best channels in channel selection by using GA.The test results of BCI Competition III Data Set I proves the superiority of the algorithm.By comparing the process of with or without feature selection,adjusting the parameters of GA,and selecting the best feature set(selecting 48.6% of the features),the classification sensitivity,specificity,precision and accuracy were 94%,98%,97.9% and 96%,respectively,which proves that using GA for feature selection improves classification performance.By comparing the algorithm with or without channel selection,the best channel combination was selected,and only 28 of the 64 acquisition electrodes were used to achieve classification sensitivity,specificity,accuracy,precision and Kappa coefficient of 98%,96%,96.08,%,97% and 0.94,respectively,exceeding the results of existing algorithms.Two methods reduced the number of features and selected the best feature sets,which improved the classification performance and shortens the classification time.It provides reference for solving redundant selection and channel problems,and provides an algorithm reference for EEG signal processing.The signal processing framework can be applied to variety of BCI systems.