speech synthesis

In PHONETICS and COMPUTATIONAL LINGUISTICS, the process of generating artificial SPEECH signals, using a MODEL of the linguistically important ACOUSTIC or ARTICULATORY properties. The devices involved are known as speech synthesizers. Acoustic domain analogs or terminal analogs replicate the acoustic properties of the VOCAL TRACT in terms of its output. The tract is represented using a SOURCE-filter model, and several devices have been devised to synthesize speech in this way, such as the early channel vocoders, the spectrogram-based FORMANT synthesizers, and LINEAR PREDICTION coefficient (LPC) synthesizers. Articulatory analogs replicate the anatomical geometry of the vocal tract between the LARYNX and the lips, in so far as information about the DYNAMIC properties of the phonatory and articulatory parameters is available. A further technique is called text-to-speech synthesis, in which written texts are automatically transformed into their spoken equivalents.