Front-End, Back-End, and Hybrid Techniques for Noise-Robust Speech Recognition

摘要: Noise robustness has long been an active area of research that captures significant interest from speech recognition researchers and developers. In this chapter, with a focus on the problem uncertainty handling in robust recognition, we use Bayesian framework as common thread for connecting, analyzing, categorizing number popular approaches to solutions pursued recent past. The topics covered chapter include 1) decision rules unreliable features model parameters; 2) principled ways computing feature using structured distortion models; 3) phase factor advanced compensation; 4) novel perspective compensation special implementation general predictive classification rule capitalizing parameter uncertainty; 5) taxonomy noise techniques two distinct axes, vs. domain unstructured transformation; 6) noise-adaptive training hybrid feature-model its various forms extension.