Methods for dynamic excitation control in parallel formant speech synthesis

Voiced excitation in the parallel formant speech synthesizer reported here uses a fixed waveform varying in fundamental period. Highly natural synthetic speech output has been achieved using time-intensive hand-copying techniques in which any dynamic changes due in natural speech to the voice source have been implemented as formant amplitude changes. Two methods of incorporating varying degrees of a dynamically varying voice source are investigated. Reliable, high-quality formant control parameters are obtained using automatic amplitude mapping and transform synthesis