Speech Synthesis for Distance Cueing in Audio Displays

The work proposed in this Phase I STTR project is aimed at developing (1) a rules-based approach for manipulating the vocal effort of a speech synthesizer suitable for text-to-speech synthesis, and (2) a non-real-time signal processing algorithm for modifying the apparent vocal effort of an arbitrary speech signal. These two technical objectives will be achieved by (1) collecting a corpus of real talkers' speech at various talker-to-listener distances, and analyzing the spectral and temporal changes that occur across those distances; (2) using synthetic speech to manipulate individual acoustic parameters likely to be cues for vocal effort changes related to distance, specifically fundamental frequency, first formant amplitude and frequency, spectral tilt, and open quotient; (3) conducting psychoacoustic testing of the real and synthetic speech to identify and rank order the importance of those acoustic cues; and (4) applying the resulting knowledge to the development of a vocoder capable of manipulating the acoustic parameters deemed as important cues for distance. Identification and rank ordering of the parameters will lead directly to synthesis and processing algorithms for manipulating the apparent vocal effort, and hence the distance, of a talker.