Acoustic Features to Discriminate between Affricates and a Fricative in Japanese

To clarify the acoustic features that distinguish the alveolo-palatal affricates [t] from the alveolar affricate [ts] or the alveolar fricative [s], Japanese words with these consonants, pronounced by single and multiple native speakers of Japanese, were analyzed with a one-third octave bandpass filter. In the frequency range of 2500–4000 Hz, [t] had much higher intensity than [ts] or [s]. In addition, when the intensity in this frequency range was used as an independent variable, a discriminant analysis showed that [t] was discriminated from [ts] or [s] with greater than 80% accuracy, but [ts] and [s] could not be discriminated. It was concluded that intensity in the frequency range of 2500–4000 Hz is a good acoustic feature to discriminate [t] from [ts] or [s] but not to discriminate between [ts] and [s].