Acoustic features of intelligible speech produced under reverberant environments

Abstract

Clear speech has been shown to have an intelligibility advantage over conversational speech in reverberant environments. Speech produced in reverberant environments is also reported to be more intelligible than ones produced in normal environments. Speakers adapt to the environments by modifying their articulation, and acoustics of speech. The speech produced under reverberant conditions should be modified similarly with the clear speech. Identifying the distinctive features which are shared between these speech types gives insights into which acoustic features are robust cues even in reverberant environments. In the present study, we investigated voices uttered by speakers who produced Japanese words under reverberant conditions (T60 = 1 s–5 s). Speech intelligibility tests, and acoustic analyses were performed. The results showed that, first, speech produced under long T60 is more intelligible in reverberant environments. Second, the intelligible speech has expanded vowel space with high F2 for front vowels and low F2 for back vowels, and high F1 for open vowels and low F1 for close vowels. Brief formant transitions were also found. These articulatory/acoustic modifications are likely to be enhancements of significant perceptual cues. The acoustic/articulatory features of each vowel become clearer (hyper-articulated speech; i.e., clear speech.) Moreover, perceptual compensation to reduce co-articulation effects should become easier.