EFFECTS OF INTER-WORD PAUSES AND SPEECH RATES ON WORD INTELLIGIBILITY IN A LONG-PATH ECHO ENVIRONMENT: FOR IMPROVEMENT OF OUTDOOR MASS NOTIFICATION SOUND SYSTEMS

摘要

Outdoor mass-notification sound systems can effectively convey emergency announcements over a wide area such that listeners would not need to use any special devices to hear the announcements. However, several factors severely degrade the intelligibility of these messages as they are being transmitted by these systems. Long-path echoes are the main causes for a lack of intelligibility of the messages transmitted by mass-notification sound systems. In view of such factors, it is important to consider effective ways of presenting speech information. In our previous studies, we investigated the effect of word familiarity and inserting pauses between the words in a speech signal. Our results clearly indicate that the intelligibility of high-familiarity words is higher than that of low-familiarity words when long-path echo conditions are present. Furthermore, the results of the experiments show that speech signals are more effectively transmitted by inserting pauses of 1 or 2 mora between words. However, regarding the length of pauses, it is unclear which is more important-the absolute length of the inserted pause or the length of the pause relative to the unit of mora. The pause length would also be affected by the speech rate, because there is a close relationship between the mora length and the speech rate. We expect that there should be optimal combinations of speech rate and pause length in long-path echo environments. Therefore, we measured word intelligibility in the presence of long-path echoes, with speech rate, pause length, and the delay time of a long-path echo as parameters. The results suggest that word intelligibility can be improved significantly by combining inter-word pauses and certain speech rates so that PSR (pause-to-speech-rate ratio) becomes an integer. This may be because, when PSR is an integer, one mora can be clearly heard/understood, facilitating the association of the presented words.