Minoru Tsuzaki
Hiroaki Kato
ATR Human Information Processing Res. Labs., 2-2 Hikaridai, Seika-cho, Soraku-gun, Kyoto, 619-02 Japan
To investigate the process of perceptual resynthesis of an ``auditory object,'' the apparent duration of a noise-replaced speech signal was measured. Subjects adjusted the duration of synthesized speech to subjectively match that of the target speech. When a portion of the target speech was replaced by a noise burst, the duration was perceived to be shorter than when no replacement was made. Besides this effect, the apparent duration of speech segments was found to be affected by their deviation from the internal template. The greater the physical duration of the target deviated from the ``naturally'' spoken token, the greater the degree of underestimation became. Interestingly, these two effects were independent of each other. This independence was replicated in the second experiment which tested wider range of durational deviations. Again, the apparent duration of noise-replaced speech was shorter than when no replacement was made. The apparent duration of gap-replaced speech was also measured in this experiment. The degree of underestimation was far greater for the latter case. These findings suggest that a shrinkage of duration occurs in the process of resynthesizing an ``auditory object'' where the process is assumed to be at a different stage than that utilizing speech-specific schema.