Making speech announcements intelligible in public spaces from a speech production view

Abstract

Speech announcements in public spaces are often hard to hear due to noise and reverberation, and this is especially true for elderly people compared to young people. This study aims to improve speech intelligibility in noisy/reverberant environments based on the way we change how we speak depending on an acoustic environment. Speech uttered in a noisy environment (noise-induced speech) is generally more intelligible for young people than speech produced in a quiet environment when both types of speech sounds are heard in noise (i.e. Lombard effect). This paper examines whether reverberation- as well as noise-induced speech are more intelligible to young and elderly people. The results of our listening tests showed that elderly listeners had significantly higher word identification scores for noise/reverberation-induced speech than speech spoken in a quiet environment. The results also showed that noise/reverberation-induced speech was more intelligible to the listeners than speech in quiet with background noise/reverberation conditions that were not only identical but also different to those used during the recording of speech. The results suggest that using noise/reverberation-induced speech for public address systems makes speech announcements more intelligible for elderly people in public spaces. [Work supported by KAKENHI (21700203) and Sophia University Open Research Center.]