Single-Channel Speech Separation Using Soft Mask Filtering

• Published in: IEEE transactions on audio, speech, and language processing Vol. 15; no. 8; pp. 2299 - 2310
• Main Authors: Radfar, M.H., Dansereau, R.M.
• Format: Journal Article
• Language: English
• Published: Piscataway, NJ IEEE 01.11.2007; Institute of Electrical and Electronics Engineers
• Subjects: Applied sciences; Detection, estimation, filtering, equalization, prediction; Estimation error; Exact sciences and technology; Filtering; Filtration; Independent component analysis; Information, signal and communications theory; Mask filtering; Masks; Mathematical analysis; Mean square error methods; minimum mean square error (MMSE) estimation; Miscellaneous; Separation; Signal and communications theory; Signal processing; Signal, noise; single-channel speech separation; Sound filters; Source separation; Spectra; Speech; Speech coding; Speech enhancement; Speech processing; Telecommunications and information theory; Vectors; Wiener filter; Performance evaluation; Signal mixing; Parameter estimation; Spectral method; Source separation; Wiener filtering; Segmentation; Acoustic signal; minimum mean square error (MMSE) estimation; Signal estimation; Modeling; single-channel speech separation; Mean square error; Mask filtering; Vocal signal; s parameter; Speech enhancement; Signal processing; Speech processing; Signal to noise ratio
• ISSN: 1558-7916; 1558-7924
• DOI: 10.1109/TASL.2007.904233

We present an approach for separating two speech signals when only one single recording of their linear mixture is available. For this purpose, we derive a filter, which we call the soft mask filter, using minimum mean square error (MMSE) estimation of the log spectral vectors of sources given the mixture's log spectral vectors. The soft mask filter's parameters are estimated using the mean and variance of the underlying sources which are modeled using the Gaussian composite source modeling (CSM) approach. It is also shown that the binary mask filter which has been empirically and extensively used in single-channel speech separation techniques is, in fact, a simplified form of the soft mask filter. The soft mask filtering technique is compared with the binary mask and Wiener filtering approaches when the input consists of male+male, female+female, and male+female mixtures. The experimental results in terms of signal-to-noise ratio (SNR) and segmental SNR show that soft mask filtering outperforms binary mask and Wiener filtering.