Dominique Salden
08-04-2008 10:31

Competitive Analysis of Voice Extraction (VE) for Speech Signal Quality Enhancement



IC Tech, Inc


Signal separation based methods have emerged as a viable means of enhancing speech signal quality. Voice extraction is a related technique based on signal separation. Unlike signal separation which decomposes the mixed signal into multiple source components, voice extraction is concerned with recovering only one signal of interest, commonly the voice of the primary speaker free of interfering signals, e.g., background noise, music, mechanical noise, and chatter of other speakers. Voice extraction results in an improved signal to noise ratio (SNR) for the speech signal of interest and improves intelligibility and other measurable criteria, such as speech recognition accuracy. Although the reach and application of voice extraction is broader than noise cancellation, at its market entry point, it competes with noise canceling products. In this paper, voice extraction is compared to existing solutions currently sold as noise canceling products on the market. Since the implementation of voice extraction embodies both hardware and software elements, comparisons are made in multiple areas, including application domains, performance, computational requirements, size, and cost.