7th International Conference on Spoken Language Processing
September 16-20, 2002
This paper investigated how Japanese listeners perceived words in a noisy environment. The data of a word identification in noise experiment reported in Yoneyama (2002) were analyzed. There are three main findings. First, more than 86% of the total responses correctly reproduced the pitch accent patterns of the stimulus words, suggesting Japanese listeners hardly misperceived pitch accent patterns of words in a noisy condition. Second, vowels were perceived more correctly than consonants. Third, MDS analyses revealed that similarities of sounds were mapped in terms of phonological features ([±high] and [±back] for vowels; [±voice] and [±sonorant] for consonants). These findings suggest that similarities of sounds that are highly related to the sonority scale of sounds in Japanese and pitch accent patterns contribute to word recognition in a noisy condition in Japanese.
Bibliographic reference. Yoneyama, Kiyoko (2002): "Similarities of words in noise in Japanese", In ICSLP-2002, 1637-1640.