Speech Digitization Fundamentals

Abstract

In digital telephony, speech is handled in digital form. Accordingly, the speech from the telephone transmitter must be converted from its analog waveform to a series of ones and zeros. A multitude of speech digitization methods has been devised; these methods can be categorized as waveform, parametric, and hybrid coding. A waveform coder takes the actual waveform and produces a series of ones and zeros representative of that waveform according to a set of rules. A parametric coder, sometimes called a source coder, attempts to detect certain characteristics of speech, such as pitch and amplitude, and produces a series of ones and zeros according to a set of rules that describes the detected speech characteristics rather than the waveform. A hybrid coder is one that uses both waveform and parametric principles in its production of a digital version of speech.