Speaker Normalization of Fricative Noise: Considerations on Language-specific Contrast

Both frication noise and vowel formants cue the place of articulation of sibilant fricatives (e.g., /s/ and /ʃ/ in English).However, only few studies have examined the effect of speaker-specific factors. This acoustic study of sibilant fricatives examines how speaker-specific formant information can improve the distinctness of two phonemic categories of sibilants: /s/ vs. /ʃ/ in French and /s/ vs. /s/ in Japanese. The results show that the center of gravity of the frication noise, normalized with respect to the subject-specific coefficient of vowel onset or vowel center formants, provides an appreciable improvement in the sibilant distinctness in Japanese. While the distinctness score of the noise is generally higher in French than in Japanese, the F2 onset patterns (/s/</s/) are throughout consistent only in Japanese, and strongly improved with a speaker normalization. These language-specific behaviors are discussed in relation with the respective phonological system.