Speech interference level as a predictor of face-to-face communication in noise

Abstract

An experiment was performed to check the validity of the distance versus speech interference level (SIL) curves for face-to-face communication in noise [Beranek, Proc. IRE 35, 880–890 (1947); Webster, Proc. Natl. Conf. Noise as a Public Health Hazard (1969)]. Monosyllabic word intelligibility was measured both with and without visual cues at talker-to-listener distances of 2, 4, and 8 ft, and at speech interference levels of 65, 75, and 85 dB. The distance-versus-SIL curves were found to be reasonably consistent with the intelligibility data when no visual cues were available, but substantially underestimated tolerable speech interference levels for barely reliable communication with visual cues. Another major source of error was the difference between the observed rate of attenuation with distance of the speech signal and the idealized, free-field rate of attenuation assumed for the distance-versus-SIL curves. Errors resulting from this assumption were found to be consequential only at the larger talker-to-listener distances. Other important factors which affected the accuracy of the curves were the total speech power of the talker, the shape of the speech power spectrum, and the rate at which the talker raises his/her voice in noise.