Vowel formant discrimination in quiet and noise for natural speech
Chang Liu, Diane Kewley-Port · The Journal of the Acoustical Society of America · 2001
Everyday communication takes place in background noise. The purpose of this study is to investigate the effects of noise on formant discrimination for naturally spoken vowels. Our analysis of formant discrimination for synthetic vowels found that formant resolution could be accounted for by excitation pattern and specific loudness models [D. Kewley-Port and Y. Zheng, J. Acoust. Soc. Am. 103, 1654 (1998)]. It is expected that excitation pattern and specific loudness model can be extended to account for the formant discrimination in natural speech in both quiet and noise, even though natural speech has more variability than synthetic speech. The present study measured formant frequency discrimination in quiet and noise for normal-hearing listeners. Thresholds were measured as ΔF in natural speech for F1 and F2 for a set of vowels /I, ε, ʌ, æ/ spoken by a female talker. Noise conditions used a long-term averaged speech spectrum to shape the noise. The signal-to-noise ratio (SNR) was manipulated systematically from −6 to +6 dB. Results showed different masking patterns for F1 and F2 with performance approaching asymptote between 0 and +3 dB SNR. Results from an analysis of these data with a specific loudness pattern model will be presented. [Work supported by NIHDCD-02229.]