Robust speech recognition by integrating speech separation and hypothesis testing

• Published in: Speech communication Vol. 52; no. 1; pp. 72 - 81
• Main Authors: Srinivasan, Soundararajan, Wang, DeLiang
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 2010; Elsevier
• Subjects: Applied sciences; Cues; Exact sciences and technology; Ideal binary mask; Information, signal and communications theory; Lattices; Masks; Mathematical models; Missing-data recognizer; Recognition; Robust speech recognition; Separation; Signal processing; Speech; Speech processing; Speech recognition; Speech segregation; Telecommunications and information theory; Top-down processing; Speech segregation; Top-down processing; Robust speech recognition; Ideal binary mask; Missing-data recognizer; Performance evaluation; Multistage method; Acoustic signal; Labelling; Frequency domain method; F region; Hypothesis test; Missing data; Accuracy; Vocal signal; Speech recognition; Speech processing
• ISSN: 0167-6393; 1872-7182
• DOI: 10.1016/j.specom.2009.08.008

Missing-data methods attempt to improve robust speech recognition by distinguishing between reliable and unreliable data in the time–frequency ( T– F) domain. Such methods require a binary mask to label speech-dominant T– F regions of a noisy speech signal as reliable and the rest as unreliable. Current methods for computing the mask are based mainly on bottom-up cues such as harmonicity and produce labeling errors that degrade recognition performance. In this paper, we propose a two-stage recognition system that combines bottom-up and top-down cues in order to simultaneously improve both mask estimation and recognition accuracy. First, an n-best lattice consistent with a speech separation mask is generated. The lattice is then re-scored by expanding the mask using a model-based hypothesis test to determine the reliability of individual T– F units. Systematic evaluations of the proposed system show significant improvement in recognition performance compared to that using speech separation alone.