Multichannel EEG Data Compression Based On Compressed Sensing

Electroencephalogram(EEG)contains a large amount of physiological and pathological information of human body.It is indispensable in the diagnosis of brain diseases and widely used in neuroscience research,physiological feedback therapy and other fields.Remote monitoring of EEG,which has an advantage of long duration,convenience and flexibility,has gradually become the trend of current technology.As the monitoring time and the number of channels increase,the amount of signal data will increase significantly,while the storage capacity and the transmission bandwidth of the portable device is limited.Therefore,it is of great significance to compress the multichannel EEG,which reduces the storage demand,the transmission bandwidth and the power consumption.Compressed sensing manages to compressed signals by converting high-dimensional signal to low-dimensional observation signal through measurement matrix.This compression method is suitable for the portable monitoring devices due to its low computational complexity.However,it is feasible to promote the compression ratio of multichannel EEG on the condition that the characteristics are taken into account.Based on compressed sensing,random signal theory and EEG characteristics,the research of multichannel EEG compression is carried out.The main research content of the paper includes:(1)On the basis of the theory and mathematical model of compressed sensing,the standard EEG database samples are reconstructed by using the common random measurement matrix in simulation.Then the optimal measurement matrix is selected according to the reconstruction error and matrix characteristics.(2)Taking the discrete cosine basis as the sparse basis,the algorithm with minimum error is selected by comparing the normalized mean square error of three reconstruction algorithms.(3)Based on the compressed sensing framework of EEG,a multichannel decorrelation algorithm is designed to reduce the correlation between the channels of EEG.Combining the decorrelation algorithm with compressed sensing and Huffman coding,a new compression reconstruction framework is constructed to further promote the compression ratio of EEG.The overall framework is coded in MATLAB,and the simulation shows that the average compression ratio increases by 9.12% compared with the compressed sensing scheme without decorrelation.(4)On the metrics of signal amplitude and operation time,the performance of the decorrelation algorithm in the case of different channel numbers and signal length is discussed.The performance of Huffman coding is also discussed according to various data sizes.Two ideas of optimization on decorrelation and multichannel reconstruction are put forward to further improve the algorithm.