Designing a quality voice: An analysis of listeners' reactions to synthetic voices

Abstract

Eight subjects listened to a set of synthetic voices reflecting a crossing of four voice qualities: head size, pitch, richness and smoothness. The listeners evaluated the voices on sixteen perceptual scales, and judged each voice's appropriateness for twenty voice-output scenarios. Factor analysis of the perceptual ratings recovered two factors, fullness and clarity. A similar analysis of the appropriateness ratings revealed three situational factors, information, entertainment and feedback. Further analyses indicated that the voice qualities associated with the three situational factors were quite different, and suggest ways to optimize voices used for a particular purpose.