Application of auditory perception in compressed sensing for speech signal

摘要

This paper reports the application of auditory perception in the sparse representation and compressed sensing for speech signal. In order to obtain a more sparse representation, an experiment of sparse coefficients' selection under masking threshold method was done comparing with the energy threshold method, results indicate that the masking threshold method is more effective. In order to verify the impact of the method to compressed sensing for speech signal, the testing speeches were compressed under compressed sensing framework by masking threshold method and energy threshold method, through some subjective and objective indicators, such as compression ratio, signal noise ratio and mean opinion score, results imply that the masking threshold method can be applied effectively in compressed sensing, and can lead an increasing of compression ratio while the hearing quality of reconstructed signal is not decreased.