الفهرس 
Speech Quality MetricsOnly 14 pages are availabe for public view
 |  | from | 16 |  |  |
| | | from | 16 | | |
Abstract
This thesis deals with the problem of blind separation of audio signals from
noisy mixtures. It proposes the application of a blind separation algorithm on the
Discrete Cosine Transform (DCT), Discrete Sine Transform (DST), or Discrete Wavelet
Transform (DWT) of the mixed signals, instead of performing the separation on the
mixtures in the time domain. Noise reduction techniques are tested as pre- and post-
processing techniques to enhance the performance of the blind signal separation
process. Both the DCT and the DST have an energy compaction property, which concentrates most of the signal energy in a few coefficients in the transform domain,
leaving most of the transform-domain coefficients close to zero. As a result, the
separation is performed on a few coefficients in the transform domain.
Another advantage of signal separation in transform domains is that the effect of
noise on the signals in the transform domains is smaller than that in the time domain due
to the averaging effect of the transform equations, especially, when the separation
algorithm is preceded by a wavelet denoising step. The simulation results confirm the
superiority of transform-domain separation to time domain separation and the
importance of the wavelet denoising step.
As a practical case, we propose the application of the different proposed
techniques of blind signal separation to improve the recognition rate of a speaker
identification system. A comparison study is held between performances of the speaker
identification systems with and without the application of the blind signal separation
algorithms in time and transform domains. The simulation results show a significant
improvement in the performance of the automatic speaker identification system with
blind signal separation.