Perceptual roles of different frequency bands in Japanese syllable identification

  • Kazuo Ueda
  • Yoshitaka Nakajima


Ueda et al. [(2010). Fechner Day 2010, Padua.] indicated that speech information could be essentially transmitted by the power fluctuations in four frequency bands. We aimed at clarifying the roles of these frequency bands in Japanese speech perception in V/CV syllable identification. We first performed factor analyses of power fluctuations of critical- band-filtered speech, and obtained four frequency bands as in the previous research. The speech was a set of V/CV patterns uttered by a male and a female speaker. The speech patterns were converted into noise-vocoded speech so that only the power fluctuation in each frequency band was preserved. There were also patterns in which one of the frequency bands was eliminated resulting in a spectral gap. Eliminating the lowest band (50-570 Hz) crucially deteriorated perceptual differentiation between voiced and unvoiced consonants. Eliminating the second lowest band (570-1850 Hz) interfered vowel identification turning almost all vowels into /i/. The roles of the other frequency bands were not obvious, but their temporal relationships with the lowest band was suggested to play a role.

Full Articles