Acoustic features of spontaneous and read conversational and clear speech produced in simulated acoustic environments

Abstract

Speech production can differ depending on how speech is elicited (e.g. spontaneous speech, read text, speaking style instructions, the speaking environment). In most studies in which different speaking styles have been elicited via instruction (e.g., clear speech) or via the speaking environment (e.g., Lombard speech), the talkers have read printed materials. There is evidence that the acoustic features of clear speech elicited by reading are similar to those observed in semi-spontaneous interaction between two interlocutors, but that clear speech changes are of a greater magnitude in the read speech than the semi-spontaneous speech [V. Hazan and R. Baker, J. Acoust. Soc. Am. 130(4), 2139–2152, 2011]. The present study examines the effects of two speaking style instructions (conversational and clear) and four simulated listening environments (quiet, 55 dB SPL of white noise, 63 dB SPL of white noise, and a reverberant environment) presented via earphones for three types of speech materials: read sentences, read passages, and spontaneously produced picture descriptions. Acoustic features relevant to clear speech and Lombard speech will be compared among the three material types.