TR-A-0084 :1990.6.28

Masato Akagi

Psychoacoustic evidence for the contextual effect model

Abstract:In previous work towards speech recognition (Akagi, 1989), a model was developed which predicted target formants in reduced vowels based on the interaction between spectral peak pairs. To substantiate this model, two psychoacoustic experiments were carried out which measured the amount of phoneme boundary shift with (1) a single formant stimulus as a preceding anchor and (2) a vowel as a preceding anchor. In the first experiment, a perceptual boundary shift with a single formant anchor was observed. When the results were compared with the spectral peak interaction obtained from real speech data using the model, this comparison showed that the perceptual boundary shift with a single formant anchor is similar to the spectral peak interaction analyzed by the model. Thus, the contextual effect between single formant stimuli should play an important role in phoneme neutralization recovery, and the neutralization recovery model is formulated as the sum of the contextual effects resulting from interaction between spectral peaks. Additionally, a comparison of these results with those of the second experiment showed that the phoneme boundary shift with a vowel anchor can be postulated as the sum of the shift with the single formant anchor and a factor from the preceding anchor. The factor can be estimated by subtracting the sum of the phoneme boundary shifts with the single formant anchors estimated by the model from the boundary shift with a vowel anchor. The difference was represented as a function of the distance between the preceding vowel anchor and the perceived vowel in a phoneme space.