Trans. Tech. Comm. Psychol. Physiol. Acoust., The Acoustical Society of Japan, Vol. 38, No. 6, H-2008-101, pp. 579-584, 2008 (in Japanese)
Perceptual speaker recognition using noise-added monosyllables
K. Amino and T. Arai
Abstract: In perceptual speaker identification, it is known that the identification accuracy depends on the stimuli presented to the listeners. In our previous experiments, the stimuli containing a nasal sound have consistently yielded better performances than the stimuli containing only oral sounds. Taking into account the fact that speaker recognition in forensics often has to do with the speech materials recorded in noise, it is necessary to investigate how the presence of noise affects the speaker identification performances. In this present study, we conducted a speaker identification experiment in an ABX paradigm where noise-added monosyllables were used as the stimuli, and investigated the effects of (1) signal-to-noise ratios, (2) the phonological contents of the stimuli, and (3) presence of the target speakers in the voice line-ups. The results showed that higher SNR and the presence of the target speaker both increased the identification accuracy. Performances on the nasals were more deteriorated than on other syllables with the line-ups where the target speaker is absent.
keywords: Speaker Identification, Noise, Individuality, Sensitivity, SNR, Nasals