Emulating Human Speech Recognition A Scene Analysis Approach to Improving Robustness in Automatic Speech Recognition

This book presents a systematic approach to the automatic recognition of simultaneous speech signals using computational auditory scene analysis. Inspired by human auditory perception, this book investigates a range of algorithms and techniques for decomposing multiple speech signals by integrating a spectro-temporal fragment decoder within a statistical search process. The outcome is a comprehensive insight into the mechanisms required if automatic speech recognition is to approach human levels of performance.