Speech Compression and Reconstruction Based on Adaptive Multiscale Compressed Sensing Theory

In this paper,a matrix form of Sym wavelet decomposition and synthesis is deduced,keeping the length of the coefficient no more than the length of original speech signals,and then we propose a framework of speech Multiscale Compressed Sensing(MCS) and an Adaptive Multiscale Compressed Sensing(AMCS) method by analyzing sparsity of different wavelet levels of speech signals.We compare AMCS with MCS by applying both methods to speech compression and reconstruction,and the reconstructed speech signal evaluated by the objective and subjective evaluation is applied to speaker recognition.The experimental results show that the reconstruction performance of speech signal based on AMCS is superior to MCS.