Techniques for Noise Robustness in Automatic Speech Recognition

Presents a comprehensive survey of the state-of-the-art in techniques used to improve the robustness of speech recognition systems Techniques for Noise Robustness in Automatic Speech Recognition will comprise chapters covering the state-of-the-art techniques written by experts in the field. Model compensation and adaptation will be considered which temporarily modify the statistical parameters employed by the recognizer to improve recognition of corrupted data. Signal compensation and signal separation methods attempt to reduce the level of noise in noisy speech. Feature compensation methods attempt to reduce the effect of corrupting noise on features derived from the speech. Missing feature methods concentrate on information-carrying components of the speech signal. Noise-robust feature estimation methods derive fundamentally noise-robust features from speech for recognition. Other methods deal optimally with packet loss, detection of bursty noises and other such phenomena. The book will comprise representative chapters that cover the state-of-the-art in all of the above approaches.* Considers the major causes of degradation in Automatic Speech Recognition (ASR) and methods for increasing ASR system robustness.* Addresses robustness issues and signal degradation - both key requirements of practitioners in ASR.* Acts as a timely exposition of the topic in light of more widespread use of ASR technology in challenging environments.* Fills a gap in the current literature as no other up to date competition exists currently.* Companion website hosting colour version of images from the book.