The Research Of Speech Coding Based On Compressed Sensing

The conventional Nyquist sampling theorem states that signal can be reconstructed without distortion only when the sampling frequency is greater than or equal to two times the highest frequency of the signal. With the amount of information requirements increasing, the signal sampling rate and processing speed are getting higher and higher, which also makes the cost of sampling increase. The theory of compressed sensing proposed in recent years completes signal compression and sampling at the same time. It makes use of the sparsity of the signal to sample at lower rate and reconstruct the signal almost without distortion,which greatly reduces the signal sampling rate,data storage and transmission costs and makes CS become a hot research field of signal processing.This paper studies compressed sensing theory and combines it with the traditional speech coding method to make CS technology used in practical communication systems. An improved PCM speech coding method based on CS is proposed, by which the measurements firstly can be coded and the original speech signal can be reconstructed from the decoded measurements. Compared with the traditional CS algorithm, the performance of reconstructed signal with this method can achieve the performance of the traditional CS algorithm.To further reduce the amount of data, a vector quantization speech coding based on CS theory is studied and proposed .The main idea of this theory is coding the overall sequence of the measurements and reconstruct the original speech signal from the decoded measurements. Simulation results show that the proposed coding method can not only achieve further compression of the amount of data based on CS, but also ensure the reconstruction performance. In addition, this paper studies and analyzes the characteristics of measurements for modeling and makes some exploration of speech parameter encoding based on CS for the future.In addition, a speech coding scheme based on wavelet packet transform and compressed sensing is proposed. Firstly, one can make two layers of wavelet packet decomposition of speech signals to get the coefficients of each node and process those nodes meeting the sparse conditions using CS with different compression ratio ,and then code them with PCM method uniformly. Those nodes that don t meet the sparse condition need be coded with vector quantization. At decoder, the corresponding decoding is completed to obtain the wavelet packet decomposition coefficients which will be used to reconstruct the speech signal. The simulation results indicate that the proposed coding method can effectively reduce the bit rate under the premise of guaranteeing the satisfied quality of the reconstructed speech signal.