A psychoacoustic method to find the perceptual cues of stop consonants in natural speech.

摘要: Synthetic speech has been widely used in the study of cues. A serious disadvantage this method is that it requires prior knowledge about cues to be identified order synthesize speech. Incomplete or inaccurate hypotheses often lead sounds low quality. In research a psychoacoustic method, named three-dimensional deep search (3DDS), developed explore perceptual stop consonants from naturally produced For given sound, measures contribution each subcomponent perception by time truncating, highpass/lowpass filtering, masking with white noise. The AI-gram, visualization tool simulates auditory peripheral processing, predict audible components sound. results are generally agreement classical studies stops characterized short duration burst followed F2 transition, suggesting effectiveness 3DDS method. However, also shown /ba/ and /pa/ may have wide band click as dominant cue. transition not necessary for /ta/ /ka/. Moreover, many contain conflicting characteristic competing sounds. robustness consonant sound noise determined intensity