Study on Joint Speech Encoding Technology based on Compressed Sensing

In this paper, a new joint speech encoding scheme based on compressed sensing was proposed. In this encoding algorithms, the compressed sensing reconstruction and PCM (Pulse Coding Modulation) were used for the speech signal encode. For the speech signal, the high frequency and low frequency coefficients can be acquired by using the wavelet transform based on the lifting scheme. For the details coefficients, using a hard threshold to remove the smaller coefficients, the high frequency coefficients are sparse, so the high frequency can be reconstructed with the compressed sensing method. Because the length of the low frequency coefficients is half of the original signal length, the PCM complex of the speech signal can be reduced. Finally, the speech can be approximately recovered with the low frequency PCM coefficients and the high frequency compressed sensing reconstructed coefficients. The experimental results demonstrate the application effectiveness for this encoding scheme in the speech processing fields.