Speech signal and source-filter model
Text-To-Speech synthesis

This approach is based on a source-filter model of speech signal production, controlled by a limited number of parameters. The synthesis consists of two stages:

• 1. using contextual rules, phonetic-prosodic information is transformed into commands for specifying the temporal evolution of synthesis model parameters;

• 2. the parameter values thus determined are used by the vocoder to synthesize the acoustic signal.

Historically, this type of technique was the first to emerge, as early as the late 1950s, with the vocoders of the time, and rules developed explicitly through the analysis of phonetic-acoustic corpora. It is still used today, after a period of eclipse in favor of unit concatenation techniques (§