Stanley M. Blaugrund
Ames Vocal Dyn. Lab., Lenox Hill Hospital, 100 East 77th Street, New York, NY 10021-1883
Kyoto Univ., Sakyo-ku, Kyoto 606, Japan
Recognition of consonants, especially voiceless plosives, is inadequate in patients who undergo multichannel cochlear implants. This study was undertaken to ascertain acoustic characteristics of voiceless plosives /p/, /t/, and /k/. Consonant-vowel syllables /pa/, /ta/, and /ka/ of a normal human voice were digitized, and processed by computer in three different ways. First, a portion of consonant signals was deleted. Second, a preceding consonant in one syllable and a following vowel in the other syllable were combined. Third, vowel /a/ was synthesized by repetition of the first cycle of following vowels, and was examined by spectral analysis. All processed sounds were perceived by five experienced listeners. Each voiceless plosive was recognized correctly by its processed syllables which had at least a 1-ms signal of consonant from the onset. Following vowels served as a cue for recognizing voiceless plosives. In following vowels, frequency information, especially in the high-frequency range, was an important factor. From these results, recognition of a short signal at the onset of preceding consonants and emphasis of high-frequency power in following vowels are essential for the improvement of speech perception of voiceless plosives in cochlear implantation.