T. V. Ananthapadmanabha
H. N. Jayasimha
Voice and Speech Systems, Temple St., Malleswaram, Bangalore 560 003, India
Klatt synthesizer (KLSYN88) is a successful hybrid formant filter model for speech synthesis [Klatt and Klatt, J. Acoust. Soc. Am. 87, 820--857 (1990)]. Several modifications are proposed to enhance the scope and quality of KLSYN88. For example, in VSS--KS, control parameters do not change abruptly; an impulse excitation is used for synthesis of stop consonants; aspirated stop consonants can be synthesized; microphonetic events with fine temporal resolution can be synthesized; parallel formant filter is correctly implemented; source--filter interaction (up to three formants) is implemented as formant frequency and bandwidth modulation rather than abrupt jumps; etc. INTANAKL estimates many of the control parameters like formant frequency, bandwidth, level, voice source parameters, aspiration noise component, pole-zero component, etc. The estimated parameters are validated by comparing the short-time spectra of the original and synthesized speech and source waveforms. VSS--KS and INTANAKL are described in this paper.