PhotoYuki Nakata
Sophia University
Pre-processing for Improving Speech Intelligibility in Reverberant Environments.

Reverberation may exist in any environment. While reverberation often enriches music, speech perception is often difficult under reverberant conditions. The purpose of our research is to improve speech intelligibility under reverberant environments by pre-processing the speech signals.

As a pre-processing technique for improving speech intelligibility in reverberant environments, "steady-state suppression" was proposed by Arai et al. (2001, 2002.). "Steady-state suppression" is a method which suppresses the amplitude of the steady-state portions of speech. The information in steady-state portions of a speech signal is relatively insignificant compared with the information in transient portions. Additionally, steady-state portions usually have more energy compared to transients. With "steady-state suppression", the energy of the speech is reduced while the important information for speech perception is retained. Decreasing the speech energy leads to a reduction of the reverberant effects.?

We conducted several perceptual experiments to investigate the effects of this processing technique. The results showed that steady-state suppression yielded significant improvements in speech intelligibility under some reverberant conditions [Arai et al., 2001, 2002.]. From further investigation, we learned that when steady-state suppression was applied to speech with decreased speaking rate, speech intelligibility increased under reverberant conditions. The maximum score of improvement by steady-state suppression was 16.6 % [Nakata et al., 2006.]. These studies showed that when we apply the steady state suppression technique between a microphone and loudspeaker, speech intelligibility is significantly improved in reverberant environments.

page top