A novel single channel speech enhancement algorithm based on sparse representation and dictionary learning

In this paper, we present a novel single channel speech enhancement method based on sparse representation and dictionary learning. In the proposed method, noise is distinguished between structured and unstructured. First, the noise dictionary is learned from a training noise database. Then, the structured noise is removed iteratively by using the noise dictionary and iterative formulas. Finally, the method adopts sparse and redundant representation over trained dictionary to extract clean speech from the unstructured noise. Extensive experimental results show that the enhanced method proposed outperforms state-of-the-art methods like multi-band spectral subtraction and the non-negative sparse coding based noise reduction algorithm.