Effect of harmonic structure of environmental noises on vowel perception
Kentaro Ishizuka, Kiyoaki Aikawa · The Journal of the Acoustical Society of America · 2001
This work reports new findings on the effect of the harmonic structure of environmental noises on vowel perception. An experiment was conducted to examine how listeners perceive natural vowels under noisy environmental conditions at around −2 dB in SNR. Vowels were presented at 60-dB SPL. Eight types of noise were used: white noise and seven types of harmonic structured noises each having the same flat spectral envelope and energy. The fundamental frequency of the vowels was about 260 Hz. A pure tone of 260 Hz and its harmonics (A), a pure tone of 520 Hz and its harmonics (B), and a pure tone of 260 Hz and its odd harmonics (C) were used as noises. Mean identification scores for the white noise and noises (A), (B), and (C) were 33.73%, 59.93%, 66.80%, and 60.27%, respectively. The spectral envelope has been widely used for automatic speech recognition as the feature parameter. However, experimental results demonstrated that perceptual identification scores are significantly different depending on the detailed spectral shape of the noise. The difference between noises (B) and (C) suggests that the even harmonic components of the vowel make less of a contribution to noisy vowel perception than the odd harmonic components.