Research On Redundant Dictionary In Voice Comprssion Sensing Application

Compressed sensing theory is a new type of signal compression contemporary theory in recent years which has been rapidly developed.Ultra-complete dictionary theory proposes a new way to build redundancy sparse signal decomposition and reconstruction.However in speech signal processing,compressive sensing applications is very rarely,redundant dictionary is still in its infancy.This article is based on a redundant dictionary application in speech signal processing for the purpose of reference to the current domestic and foreign redundant dictionary in speech signal processing applications,propose a new wavelet-based orthogonal base to build redundant dictionary,application cascade sparse decomposition of the speech signal,and the experimental validation of the simulation.This paper proposed algorithm respect,implementation,improved algorithm of these three aspects described in detail,and its main contents are:1.algorithm proposed:the relationship between the latest ultra-complete dictionary database and wavelet frame theory,we propose the use of multiple sets of orthogonal wavelets to build redundant dictionary thoughts and constructed according to the principles of orthogonal wavelets,as well as non-coherent dictionary theoretical construct redundant dictionary;2.implementation of the algorithm:the voice signal on a redundant dictionary application proposed cascade sparse decomposition,and put forward a feasible method of the signal and the dictionary of atoms matching,build system block diagram of the completion of the algorithm design;3.Improved Algorithm.According to the principle of the experimental data recording and analysis on some module the paper improve the algorithm so that it can perfect the speech signal sparse decomposition.Simulation results show that the speech signal processing algorithm on sparse decomposition which significantly effect the compression ratio of 0.8 can reconstruct the speech signal very well,and the transmission of voice and reaches the minimum standard communication requirements.