Acoustic features correlated to perceived urgency in evacuation announcements

•We examined the acoustic features controlling the perceived urgency in speech with kept clarity.•We used converted speech stimuli that replaced the three acoustic features,i.e., speech duration, fundamental frequency (F0), and spectral sequence (voice intensity), with each other among the five speech stimuli, which were spoken by a professional announcer in a real disaster.•The results of experiments showed that the perceived urgency was most influenced by F0, the F0 average and F0 dynamic pattern, while kept clarity of speech. To encourage prompt evacuation behavior during disasters, evacuation announcement systems are required to adjust the perceived urgency of announcements to the danger of a situation while maintaining voice clarity. In this study, we aimed to understand acoustic features correlated to the perceived urgency in speech when clarity is maintained. For this purpose, we used a speech synthesis tool to manipulate the key acoustic features: the time duration, F0 (instantaneous fundamental frequency), and spectrum. Specifically, we replaced these features in five real evacuation announcements that were spoken clearly by a TV announcer during an actual disaster and had different magnitudes of urgency. We found quantitatively and qualitatively that the perceived urgency was mostly influenced by the F0 of speech. Furthermore, by manipulating the F0 time average and variation, we compared the influence of the F0′s static constant feature and its dynamic fluctuation pattern on the magnitude of perceived urgency. The results indicated that both types of F0 information influenced the magnitude of perceived urgency in real Japanese speech. Our results suggest that the sense of urgency in evacuation announcements can be controlled by adjusting the F0 while maintaining voice clarity.