STRAIGHT: A new speech synthesizer for vowel formant discrimination

Abstract

The present study investigated whether a new tool for nearly natural speech synthesis, STRAIGHT [Kawahara et al., Speech Commun. 27, 187–207 (1999)], could be used for fine manipulation of vowel formants, using a psychophysical test of formant discrimination. Thresholds for formant discrimination of F1 and F2 for an /ɛ/ vowel, originally synthesized by the KLTSYN [Klatt, J. Acoust. Soc. Am. 67, 971–995 (1980)] and then resynthesized by STRAIGHT, were estimated. Thresholds for vowels generated by KLTSYN and by STRAIGHT were not significantly different. This result validates that STRAIGHT resynthesis can finely manipulate formant frequencies from natural speech for use in speech perception experiments.