Event Details
Subspace Speech Enhancement
Presenter: S.-H.Yu
Supervisor:
Date: Wed, December 11, 2002
Time: 14:00:00 - 15:00:00
Place: EOW 230
ABSTRACT
Abstract:
Speech enhancement attempts to improve some aspects of speech corrupted by noise for the human listener or to improve the speech signal so that it may be better exploited by other speech processing algorithms. The purpose of this work is to examine the potential of using a signal subspace approach for speech enhancement. This includes development and analysis of algorithms originating in the area of numerical linear algebra.
The challenge is to estimate a clean signal based on a single noisy signal. It is approached initially by performing an analysis of subspace principles, then it is applied to speech signals in order to characterize the usefulness of defining a signal subspace. The theory is formulated by means of the SVD (singular value decomposition) and QSVD (quotient singular value decomposition). Finally, a number of demonstrative examples are also given.
Free and open to the public