Speech production in amplitude-modulated noise

Abstract

The Lombard effect refers to the phenomenon where talkers automatically increase their level of speech in a noisy environment. While many studies have characterized how the Lombard effect influences different measures of speech production (e.g., F0, spectral tilt, etc.), few have investigated the consequences of temporally fluctuating noise. In the present study, 20 talkers produced speech in a variety of noise conditions, including both steady-state and amplitude-modulated white noise. While listening to noise over headphones, talkers produced randomly generated five word sentences. Similar to previous studies, talkers raised the level of their voice in steady-state noise. While talkers also increased the level of their voice in amplitude-modulated noise, the increase was not as large as that observed in steady-state noise. Importantly, for the 2 and 4 Hz amplitude-modulated noise conditions, talkers altered the timing of their utterances, reducing the energetic overlap with the masker by approximately 2%. However, for the 1 Hz amplitude-modulated condition, talkers increased the overlap by approximately 4%. Overall, the results demonstrate that talkers are sensitive to the temporal aspects of noisy environments and will alter their speech accordingly.