Poster D88, Monday, March 26, 8:00-10:00 am, Exhibit Hall C
The effects of the amplitude envelope of speech on speech intelligibility
Mako Ishida1,2,3, Takayuki Arai3, Makio Kashino1; 1NTT Communication Science Laboratories, 2Japan Society for the Promotion of Science, 3Sophia University
The current study examines the intelligibility of speech by degrading speech signal by modifying the amplitude envelope of speech in two different ways. This study examines how our brain decodes spoken messages from degraded speech signal. In Experiment 1, participants listened to locally time-reversed speech where every certain length of speech signal was reversed on the time axis. Speech was locally time-reversed at every 10, 30, 50, 70, 90, or 110 ms. In Experiment 2, participants listened to speech in which the amplitude envelope of speech signal was modified by filtering modulation frequency components. The cut-off frequency was set as 1, 2, 4, 8, 16, or 32 Hz. The manipulation of local time reversal as well as modulation frequency filtering induced the change of amplitude envelope of speech from the gradual to the drastic level. Participants were native English speakers in the United States, and they transcribed English sentences they heard through headphones. The results suggest that speech is intelligible when the reversed segment length is relatively short, and when the modulation frequency components are relatively preserved with higher cut-off frequency. However, speech becomes gradually unintelligible when the reversed segment length exceeds around 50-70 ms, and when the cut-off frequency is lower than 4-8 Hz. The results of two different experiments suggest that lower modulation frequency is critical for speech intelligibility and perceptual reconstruction of speech. It seems to suggest, that speech is decodable, by humans, when the lower modulation frequency components are preserved in the amplitude envelope of speech.
Topic Area: PERCEPTION & ACTION: Audition