Perceived urgency of synthesized voice warnings

Abstract

Speech synthesis has been used successfully in a number of important applications such as the generation of warning messages in high-workload environments. In this thesis, the effects of synthesized voice warning parameters on perceived urgency were examined in order to build a detailed and usable description of the relation between parameters of synthesized voice warnings and perceived urgency. Ten native and ten non-native English speakers participated in four experiments to evaluate and quantify the effects of the voice parameters. The first experiment was designed to identify the voice parameters influencing perceived urgency. The results showed that speech rate, average fundamental frequency ($F_0$), voice type, and fundamental frequency contour have clear effects on the perceived urgency of synthesized voice warnings. The second and third experiment deals with the quantification of the relation between quantifiable parameters and perceived urgency. The effects of average $F_0$ and speech rate on perceived urgency were scaled using an application of Stevens``s power law. In addition, the results showed significant differences in perceived urgency of average $F_0$ and $F_0$ contour types between native and non-native English speakers. In the final experiment, detailed analysis of the relationship between various $F_0$ contour types and perceived urgency were conducted. The results of this study would be useful for design of the new warning signals and to improve existing warnings of industrial, military and public applications.