Multiple Audio Source Separation by Using Intra-Object-Sparsity Encoding Framework

Jundai Sun,Maoshen Jia,Changchun Bao,Boxuan Song

DOI: https://doi.org/10.1109/icspcc.2017.8242371

2017-01-01

Abstract:This paper proposes a multiple audio source separation method by using the intra-object-sparsity (in each frame, the energy of an audio signal concentrates on small number of time-frequency instants) encoding framework. Specifically, by applying the intra-object-sparsity of audio signal, each source is encoded to obtain a sparse representation of it while preserves the major energy of the original signal. Since, most of the multiple source separation algorithms for speech sources can be extended to the audio sources. The combination of the intra-object-sparsity encoding framework and source separation method can effectively eliminate the cocktail party problem which lead to bad separation quality. The evaluations reveal that the proposed method achieves a higher separation quality compared with the existing techniques and robust over different types of audio signals.