Research On Reconstruction Algorithm Of Local Field Potential Based On Compressed Sensing

ObjectiveBased on the traditional sampling theorem, the sampling of Local field potentials(LFP) will produce an enormous amount of data which will bring great pressure on data storage and transmission. To reduce the sampling rate and sampling data, a new method based on compressed sensing to reconstruct LFP signal was proposed. Through quantitative analysis of computational complexity and reconstruction accuracy of different combinations of sparse decomposition and reconstruction algorithms for compressed sensing to find out the suitable compressed sensing method for LFP signal.Contents1. Simulation experiment of applying compressed sensing to sparse signalThis article did simulation experiment of sparse signal reconstruction based on compressed sensing. Identity matrix was chosen as the sparse matrix, and orthogonal matching pursuit algorithm was chosen as the reconstruction matrix to reconstruct the signal with sparsity of10and100.2. Research on reconstruction algorithm of Local Field Potential based on compressed sensingUsing three sparse matrix and two reconstruction algorithm to achieve the LFP signal reconstruction based on compressed sensing. The sparse matrix includes discrete cosine matrix (DCT), discrete Fourier matrix (DFT) and discrete wavelet matrix (DWT); The reconstruction algorithm includes Basis Pursuit (BP) and orthogonal matching pursuit (OMP).3. Performance evaluation of compresses sensingWe chose four evaluation factors to compare and analyze reconstruction accuracy and computational complexity of the compressed sensing implementation.Results1. Through the experimental comparison, we can find that:when the signal length N was1024, the sparsity was10, then measurements of40will be able to reconstruct the original signal;when the sparsity was up to100, the reconstruction was good if measurements was300and it will had better reconstruction effects as the measurements increased.2. When chose DFT matrix, DWT matrix and DCT matrix as a sparse matrix, it can better achieve LFP signal sparse decomposition.3. When using OMP recovery algorithm to achieve reconstruction of the LFP signal, the three sparse matrix had a stable effect on the reconstruction; DCT sparse matrix and DFT sparse matrix both had a smaller reconstruction error., and the matching degree is up to0.97or so. When chose DFT matrix as the sparse matrix, the signal reconstruction is better by analysis of variance. But the reconstruction time of DFT sparse matrix is twice of the other two sparse matrix.4. When using BP recovery algorithm to achieve the reconstruction of the LFP signal, DCT sparse matrix can reconstruct the original signal exactly, the matching degree is up to0.98. Compared to the OMP reconstruction algorithm, BP reconstruction algorithm has more computational complexity and requires more computational time.Conclusion1. As a compression technology, the compressed sensing theory applied in LFP signal has its application binding, and it relies on the actual application requirements and acceptable error range. When selected OMP as reconstruction algorithm, DCT and DFT as sparse matrix, compressed sensing theory can realize LFP signal reconstruction well. When elected DFT as the sparse matrix, the reconstruction effect was better.2. Selected BP as reconstruction algorithm, DCT as sparse matrix, reconstruction accuracy was much better but it required more computational time.3. Reconstruction of LFP signal base on compressed sensing can significantly reduce the LFP signal sampling rate, and it can reconstruct the original signal when the sampling rate is in half.