Motion Imagine Eeg Recognition Algorithm Based On Sparse Representation

In recent years, brain-computer interface has become one of the research hotspots infields of biomedical and signal processing. It is a new method which human use tocommunicate with the outside world. This method can send the messages of brain tocontrol the external devices directly, without peripheral nerve and muscles. There aremany brain-computer interface systems with different function have been developed basedon this characteristic. Such as, we can use signals of brain to control the wheelchair. Thesesystems can convenient human’s life. To obtain a better brain-computer interfacesystem,we should improve the recognition rate of brain signals. In order to achieve thisgoal, a lot of experiments have been made by the domestic and international researchers,and many achievements have been made. In this paper, we conducted in-depth of EEGabout the acquisition of EEG, preprocessing, feature extraction and classification fouraspects. This paper will raise two new sparse representation algorithms for EEG signalsbased on the analysis and summary of the domestic and international relevant researchresults.Firstly, considering the importance of the local information of signal, we use amethod of sub module of signal sparse representation for identification. In view of thesignal which containing noise, preprocessing with band-pass filter is proposed in thispaper. Then we process the filtering of subsidiary module with the common space mode,and extracting their energy feature. And then we represent the data of subsidiary modulesparsely. Then the sub module’s category is judged by the sparse approximation residual.Each sub module has a ticket, and calculates all sub module of each type. Then the testsample is classified to the kind which has the most votes.Secondly, in order to improve the classification accuracy of brain-compu-ter signal, a kind of sparse representation classification algorithm-based on projectionpower features is proposed in this paper. The first step is filtering the signals. Then we usea sliding window function that can divide the signal into several sections. We extract thefeatures of every segment of a signal by a method-based projection power features. Thefeatures of each segment of a signal are combined into a feature vector. We receive the features of the signal, then we represent the features of the signal sparsely. Finally, weuse the results of sparse approximation residuals to determine the class which the testsignal is belonged.Finally, we summarize this article, and we are prospected the future of EEG.