Proc. of the International Symposium on Room Acoustics: Design and Science, Hyogo, 2004

The effect of pre-processing for improving speech intelligibility in the Sophia University lecture hall

N. Hodoshima, T. Goto, N. Ohata, T. Inoue and T. Arai

Abstract: One reason reverberation degrades speech intelligibility in an auditorium is the effect of overlap-masking, which occurs when segments of an acoustic signal are affected by reverberation components of previous segments (Bolt and MacDonald, JASA, 1949). Arai et al. suggested a pre-processing technique to prevent the effect of overlap-masking. Their technique champions the suppression of steady-state portions of speech that have more energy but that are less crucial for speech perception than are transitional portions of speech (Arai et al., Acoust. Sci. and Tech., 2002). Arai et al. confirmed promising results for improving speech intelligibility. Hodoshima et al. explored the effect of steady-state suppression with a set of artificial impulse responses and showed that steady-state suppression is an effective pre-processing method to prevent degradation of speech intelligibility under specific reverberant conditions (Hodoshima et al., Acoust. Sci. and Tech., 2004).
In this study, we conducted a perceptual test using Arai's pre-processing technique in an actual environment: the largest lecture hall in Sophia University in Tokyo, Japan, which has a reverberation time of approximately 1.0s. This study confirmed that steady-state suppression is an effective pre-processing method for improving speech intelligibility not only in simulated reverberant conditions but in an actual hall.

Keywords: Reverberation, Overlap-masking, Speech intelligibility, Steady-state suppression

[PDF (222 kB)]