基于小波变换的电生理信号压缩采样与重构  

Compressed Sampling and Reconstruction of Electropysiological Signals Based on Wavelet Transformaion

作　　者：刘欣阳[1,2,3] 徐声伟[1] 周帅[1,3] 宋轶琳[1] 罗森林[2] 蔡新霞[1,3] 

机构地区：[1]中国科学院电子学研究所,北京100190 [2]北京理工大学信息与电子学院,北京100081 [3]中国科学院大学,北京100049

出　　处：《网络新媒体技术》2014年第2期7-13,共7页Network New Media Technology

基　　金：中国科学院战略性先导科技专项课题资助项目(XDA06020101)

摘　　要：压缩感知是一个新兴领域,该理论可对信号以低于奈奎斯特采样率的速率进行成比例压缩采样,用来降低数据存储。本文基于压缩感知和小波变换,设计并实现了神经动作电位信号的压缩与重构。首先在小波域构造了64位神经动作电位信号的稀疏矩阵,然后设计了64位神经动作电位信号的2∶1压缩矩阵与OMP(Orthogonal Matching Pursuit)重构算法,并通过编程仿真实现,可以完成信噪比较高的压缩信号的高精度恢复。仿真结果表明,重构信号与原信号的关键值相对误差小于15%。CS（ Compressed sensing） is an emerging field. The theory acquires data from known signal structure at a rate proportional to the information rate rather than the Nyquist rate, so that the amount of data storage can be reduced. Neural Signal Compression and Re- construction system, which is based on compressed sensing and wavelet transform, is designed and realized in this paper. First of all, a sparse matrix of 64 bits neural signals is constructed in the wavelet domain. Secondly, a programme is made to realize the compres- sion of the 64 bit neural spike signal, which has high SNR, in the ratio of 2 to 1, and to reconstruct it accurately by using OMP （ Or- thogonal Matching Pursuit） algorithm. The simulation results show that all key values of the relative error between original signals and reconstructed signals are below fifteen percent.

关 键 词：压缩感知 小波变换 稀疏表示 OMP算法 

分 类 号：TN911.7[电子电信—通信与信息系统] O174.2[电子电信—信息与通信工程]