Synthesis of a female voice, a preliminary study

Abstract

The ultimate aim of my work is to synthesize a female voice with the help of our OVEIII synthesizer. I have started this work by surveying the existing literature where only three articles describing experiments in this area could be found, none of which claimed to have been successful. We copied one of those experiments and also tried to copy one utterance made by a female speaker. None of these synthesized samples were acceptable as a female voice. To gain more knowledge about the perceptual load of different parameters we ran a test where the fundamental frequency, the bandwidths, and the higher formants were varied. The listeners seemed to be aware of differences in average fundamental frequency that were greater than 10 Hz and were more sensitive to changes in the first formant than in the higher formants.