ABSTRACT
The second area of speech technology applications covered in the final Part IV of this book is speech enhancement. Speech enhancement refers to the restoration of clean speech, either in the form of speech waveforms for enhanced human perceptual listening or in the form of speech features for enhanced or robust speech recognition, starting from speech corrupted by distorting acoustic environments. This distortion may be due to additive ambient noise, linear or nonlinear channel (convolutional) distortion, or interfering speech. As in ASR, the dynamic modeling and optimization principles as main threads throughout this book are also clearly illustrated in speech enhancement applications, to which we devote this chapter.
