The effect of a preprocessing approach improving speech intelligibility in reverberation considering a public-address system and room acoustics

Abstract

This study evaluates a preprocessing approach for reducing reverberation effects when the input signal is not ideal, dry speech, but rather a realistic speech signal picked up by a close microphone in a room. And this study shows how the situation affects the input and the further approach compared to a dry signal as the input. Steady-state suppression, as described by Arai et al. [Acoust. Sci. Technol. 23, 229–232 (2002)], was used as a preprocessing approach that processes a speech signal before it is radiated from loudspeakers. A lecture was simulated in two different halls (reverberation times of 1.2 and 1.8 s) in which public address systems were installed. The simulation software CATT-Acoustic was used and impulse responses were calculated for the input to the preprocessing approach and for a listener position. Stimuli for a syllable identification test were prepared by convolving speech signals with the calculated impulse responses. Speech signals were given with and without steady-state suppression. The inclusion of natural and electroacoustical impulse responses makes the study of steady-state suppression more realistic and tests its robustness. [Work supported by JSPS (176911).]