Japanese listeners' perception of English fricatives in AMR-NB cell phone speech

From the fact that Japanese listeners have difficulty perceiving English fricative sounds in natural speech, it is natural to expect that such listeners would have even more difficulty perceiving fricatives in degraded speech. This research tests the ability of 72 Japanese university students to identify fricatives in English words presented in cell phone speech simulated with the AMR-NB codec. Participants were presented with 120 stimuli – 40 at CD quality speech (44.1 kHz), 40 downsampled to 8.0 kHz, and 40 resembling cell phone speech. There were 20 words, each with a fricative consonant (/s/, /S/, /f/, /T/, /z/, /dZ/, /v/ or /D/) in either word-initial or word-final position (e.g., “sick” or “mouse” for /s/). In addition, words were presented in two different contexts: in isolation and embedded in a carrier sentence. Results showed that the ability to correctly identify the fricatives depends not only on signal type, but also on the type of fricative and whether it is located at the beginning or end of the word. We found that voiceless sounds are more often misidentified than voiced sounds, in most contexts and across most signal types. We also found that having the stimulus word appear at the end of a carrier sentence helps perception of the fricative sound in that stimulus. An interesting point is, perceptibility of /T/ as in “thick” was not significantly better than chance in both types of speech signal (normal and cell phone). It can be presumed that the reason for this is the Japanese language does not contain the sound /T/. Correlation of perceptibility with duration of voiceless fricatives and perceptibility with mean intensity are also described in this paper to determine what might be affecting the perceptibility rate.