Speech Signal Processing and Simulation Analysis Based on Compressed Sensing

Abstract

This paper applies the compressed sensing theory to the reconstruction of speech signal. It makes deep study on speech signal, including sparsity and measurement matrix on different bases, and the influence on reconstruction effects. According to the principle theory and necessary conditions of measurement matrix this paper put forward an improved scheme. For the problem in slow reconstruction speed and given iteration numbers, we adopt the idea of selection criteria for matching atoms and backward projection in optimal orthogonal matching algorithm to perform iterative reconstruction. By experiments compared to original algorithms, the improved scheme can obtain accurate reconstruction signals and better performance, reducing the number of sampling points simultaneously.