Preprocessing effects on speech intelligibility in reverberation using mixed natural and electroacoustical sounds

In Proc. of the Japan-China Joint Conference of Acoustics, Sendai, 2007

Preprocessing effects on speech intelligibility in reverberation using mixed natural and electroacoustical sounds

N. Hodoshima, P. Svensson and T. Arai

Abstract: This study evaluates a preprocessing approach for reducing reverberation effects when listeners hear both sounds from a talker and from a loudspeaker. Steady-state suppression, as described by Arai et al. [Proc. Autumn Meet. Acoust. Soc. Jpn., 2001; Acoust. Sci. Tech., 23, 229-232, 2002], was used as the preprocessing approach which suppressed steady-state portions of a speech signal before it is radiated from loudspeakers. We simulated two different halls in which public address systems were installed. Stimuli for a syllable identification test were prepared by convolving unprocessed and steady-state suppressed signals with calculated impulse responses, and were presented to 20 young adults with normal hearing. Results showed that a loudspeaker gain affected the performance of steady-state suppression. Results also showed that the effect of the mixture of sounds from a loudspeaker and from a talker on the performance of steady-state suppression was negligible when a direct-to-reverberation ratio is high at a talker microphone. The mixture of natural and electroacoustical sounds makes the study of steady-state suppression more realistic and tests its robustness.

[PDF (163 kB)]