A hybrid time-frequency domain articulatory speech synthesizer

High quality speech at low bit rates (e.g., 2400 bits/s) is one of the important objectives of current speech research. As part of long range activity on this problem, we have developed an efficient computer program that will serve as a tool for investigating whether articulatory speech synthesis may achieve this low bit rate. At a sampling frequency of 8 kHz, the most comprehensive version of the program , including nasality and frication, runs at about twice real time on a Cray-1 computer. L INTRODUCTION OW bit rate coding of speech (Flanagan et al. [l]) is an important objective of current speech research. There are essentially three different methods. used for speech synthesis from low bit rate data (e.g., Flanagan [2] and Linggard [3]): formant synthesis, synthesis from linear prediction coefficients (LPC), and articulatory speech synthesis. Formant synthesis models the spectrum of speech while linear prediction models the signal wave-form using correlation techniques. For these methods there exist accompanying analysis procedures for obtaining the low bit rate data directly from the speech input. Articulatory synthesis models the speech production mechanism directly. At present, however, there is no appropriate analysis procedure which satisfactorily solves the related " inverse " problem of obtaining .articulatory parameters from spoken utterances (for preliminary at