Blind separation of speech mixtures via time-frequency masking
IEEE Transactions on Signal Processing
| Hi-index | 0.08 |
In this paper, the DUET binaural model is extended to the single-channel mixing model where only one microphone is available for recording. A novel ''artificial stereo'' mixing model is proposed to create a synthetic stereo signal by weighting and time-shifting the original single-channel mixture. Separability analysis of the proposed model has also been derived to verify that the artificial stereo mixture is separable. This work, therefore, relaxes the under-determined ill-conditions associated with monaural source separation and path the way for binaural source separation approaches to solve monaural mixture.