Abstract
Noises with different frequency bands, periods of duration and relative intensity levels to the following vowels were combined with artificial vowels to synthesize artificial voices. The regions of Japanese voiceless consonants were roughly determined through the hearing tests of these artificial voices by 9 skilled and 86 unskilled peoples. In general the regions of [ts], [t⎰] and [t] were observed as distributed in higher frequency bands than that of [k] and [p]. As to the products of intensity levels and periods of duration, [t] and [p] had smaller values on them than [ts], [t⎰] and [k]. Fricative voices were heard when noises were made to build up gradually and the periods of duration were made long, in which case, at higher frequency bands they were heard like [s], while at lower bands like [h]. On the other hand, the effect of following vowel on the regions of each consonants, correlation between intensity and period of duration, tendency of confusion and the conditions for sounding like sonants and fricatives were studied through the results. As to the personal differences, they were divided into 3 types and the distribution were also studied.
