ABSTRACT

The preceding two chapters were concerned with analysis and processing of deterministic signals, where we managed to dispense with the notion of uncertainty or randomness in the signal description. In the real world we live in, however, experiences of randomness permeate our daily lives. This could be either due to the true nature of our encounters, or due to our admitted lack of understanding of their character, or even due to our deliberate attempt to use randomness (in spite of the understanding) to our advantage. In speech processing, all these three possibilities are common. In particular, the philosophy of building random, statistical, "ignorance' models for the speech data has been prevalent and rooted in much of modern speech processing techniques. We will discuss many such techniques in the remainder of this book.