Adaptive speech enhancement using sparse prior information

In recent years, sparse representation is adopted to improve the quality of noise corrupted speech. However, the representation of noise is also found to be sparse in some special cases, which degrades the performance of sparsity based speech enhancement. An adaptive speech enhancement algorithm using sparse prior information is proposed in this paper. In the proposed method, speech enhancement is casted to an optimization problem, where linear prediction (LP) residual and DCT coefficients are combined and adopted as the representation of speech to ensure that noise is dense in the such domain. Other features, including speech energy, noise energy, and interframe correlation are also considered as constraints to improve the quality and intelligibility of recovered speech. Experiment results show that the proposed algorithm exceeds the reference algorithms in various noise scenarios, especially, in the cases of narrowband noise and low SNR.