Technical Report of IEICE Japan, Vol. SP2004-154, pp. 1-6, 2005 (in Japanese)
Evaluation of steady-state suppression of speech for elderly people in reverberant environments
K. Yasu, Y. Miyauchi, N. Hodoshima, N. Hayashi, T. Inoue, T Arai and M. Shindo
Abstract: In reverberant environments, speech intelligibility is reduced and it is difficult for people to perceive speech. This is due to overlap-masking by reverberation components of previous segments (Bolt et al., 1949). For elderly people and hearing-impaired people, it is more critical issue to perceive and to understand speech in reverberant environments (Fitzgibbons and Gordon-Salant, 1999). There are two general approaches for improving speech intelligibility in reverberant environments: pre-processing and post-processing. To obviate the deterioration of speech intelligibility, Arai et al. suppressed steady-state portions of speech (steady-state suppression) that contained more energy compared with transitions but which were less crucial for speech perception, and confirmed a promising result for improving speech intelligibility (Arai et al., 2002). In previous studies, the effect of steady-state suppression was evaluated by measuring speech intelligibility of original and steady-state suppressed speech in simulated reverberation environments by convoluting speech with an impulse response in an auditorium at a sound proof room (Hodoshima et al., 2003, 2004). The results of the experiments with young normal hearing people showed that there were significant improvements in speech intelligibility using steady-state suppression in reverberant times between 0.7 s and 1.2 s. In this research, we conducted an experiment for evaluating steady-state suppression using reverberation times of 1.0 s and 1.3 s with fifty elderly people. We found that there were significant improvements in both reverberant conditions. Also, steady-state suppression yielded more improvement in intelligibility for a presbycusis group than a normal hearing group.
Keywords: speech enhancement，elderly people，reverberation，steady-state suppression，speech intelligibility