Bøker av Raghunath Holambe

Automatic speaker recognition has been an active research area for more than four decades and the technology has gradually matured to a state ready for real life applications. However, contamination by additive noise becomes a major obstacle to commercialize the use of this technique. A more realistic way to try to improve on the situation is to search for ways to improve the robustness of the features used. Therefore, this book addresses the issue of speaker identification under mismatched conditions. The main focus of the book is on the low-level spectral features. This book covers the feature extraction process, including the speech production and perception systems. The importance of high frequency components and the phase information (which is normally discarded by most of the state-of-the art features) are investigated and different noise robust algorithms are proposed which makes use of wavelet packet transform, Teager energy operator and AM-FM model. This book would help the researchers and engineers to grasp the basics and quickly move on to more sophisticated techniques.