論文

The perception of lexical pitch accent in Japanese was assessed using noise-excited vocoder speech, which contained no fundamental frequency (fo) or its harmonics. While prosodic information such as in lexical stress in English and lexical tone in Mandarin Chinese is known to be encoded in multiple acoustic dimensions, such multidimensionality is less understood for lexical pitch accent in Japanese. In the present study, listeners were tested under four different conditions to investigate the contribution of non-fo properties to the perception of Japanese pitch accent: noise-vocoded speech stimuli consisting of 10 3-ERBN-wide bands and 15 2-ERBN-wide bands created from a male and female speaker. Results found listeners were able to identify minimal pairs of final-accented and unaccented words at a rate better than chance in all conditions, indicating the presence of secondary cues to Japanese pitch accent. Subsequent analyses were conducted to investigate if the listeners' ability to distinguish minimal pairs was correlated with duration, intensity or formant information. The results found no strong or consistent correlation, suggesting the possibility that listeners used different cues depending on the information available in the stimuli. Furthermore, the comparison of the current results with equivalent studies in English and Mandarin Chinese suggest that, although lexical prosodic information exists in multiple acoustic dimensions in Japanese, the primary cue is more salient than in other languages.

The perception of Japanese pitch accent was investigated using synthetic speech in which the F0 was artificially removed. While the F0 is said to be the primary cue to pitch accent in Japanese, it is not certain whether acoustic correlates other than F0 exist. The results of previous production studies that examined vowel duration or intensity as a correlate of pitch accent are mixed. The present study attempts to find correlates of pitch accent from the other end, i.e., perception. A native Tokyo Japanese speaker produced 14 disyllabic minimal pairs that differed only in the presence or absence of accent (e.g./haná/'flower' when accented vs./hana/'nose' when unaccented) in a carrier sentence. The utterances were then edited by replacing the F0 with random noise, which is somewhat similar to whispered speech in that there is no voicing. Twenty-two native speakers of Tokyo Japanese identified the words in two kinds of stimuli, the 14 minimal pairs as produced by the speaker and the synthetic speech. The results suggest evidence of pitch accent in synthetic speech. Furthermore, exposure to natural speech seems to improve the listener's ability to identify words in subsequent synthetic speech and vice versa, to varying degrees.

Formant frequencies were examined as a possible acoustic correlate to Japanese accent. A perception study conducted by the first author of the present study used synthetic speech stimuli in which the harmonic structure of each spectrum (F0) was removed from speech produced in a normal manner and replaced by white noise. The stimuli created this way ensured that the only property altered from the normal speech would be the presence or absence of the F0. The result found that Japanese listeners reliably identified minimal pairs of words that differed only by accent, indicating that F0 is not the only acoustic cue to Japanese accent. Since previous studies on whispered speech report a positive correlation between the pitch height intended by speakers and formant frequencies, formant frequencies were measured as a possible correlate to accent. However, the analysis did not find any correlation between the F0 movements observed in the original normal speech and the formant frequencies measured in the synthetic speech. This result suggests that acoustic properties other than the F0 are also affected in whispered speech and claims made about the positive correlation between the intended pitch and formant frequencies in whispered speech do not hold in normal speech. © 2013 Acoustical Society of America.

Phonological contrasts are typically encoded with multiple acoustic correlates to ensure efficient communication. Studies have shown that-such phonetic redundancy is found not only in segmental contrasts, but also in suprasegmental contrasts such as tone. In Japanese, fundamental frequency (F0) is the primary cue for pitch accent. However, little is known about its secondary cues. In the present study, a perception experiment was conducted to examine whether any secondary cues exist for Japanese accent. First, minimal pairs of final-accented and unaccented words were identified using a database, resulting in 14 pairs of words. These words were then produced by a native Tokyo Japanese speaker, and presented to participants in both unedited and edited forms. Edited speech stimuli were created by replacing F0 in the natural speech stimuli with white noise. While word identification by Tokyo Japanese speakers had higher accuracy for natural speech than for edited speech, the accuracy exceeded the chance level for edited speech, suggesting the existence of secondary cues for Japanese accent. Acoustic analysis of the stimuli revealed that relative mean amplitude and relative maximum amplitude were greater for final-accented words than for unaccented words. (C) 2017 S. Karger AG, Basel

Lexical pitch accent in Japanese is primarily realized as a steep fall in from an accented syllable into the following one. In addition, when a phrase that contains an accented syllable is followed by another phrase, the following phrase undergoes downstep, a compression of the range. Furthermore, while their acoustic identity is not yet clear, secondary cues to Japanese pitch accent are known to exist. The present study examined how speakers of Tokyo Japanese used acoustic information from these three sources in perceiving lexical pitch accent in Tokyo Japanese. Listeners heard stimuli in which the acoustic cues related to accent were independently manipulated and were asked to identify if a word presented sentence-medially was a final-accented word or its unaccented counterpart. Results found that listeners' judgments of words were most consistent with the presence or absence of downstep. That is, listeners identified that the preceding phrase contained an accented word when the following phrase was downstepped. Listeners also used the fall to determine if the word in question was a final-accented word or an unaccented word. Secondary cues to pitch accent were most weakly related to listeners' identification of accent.