A Japanese TTS system based on multiform units and a speech modification algorithm with harmonics reconstruction

This paper proposes a new text to-speech (TTS) system that utilizes large n umbers of speech segments to produce very natural and intelligible syntheti c speech. There are two innovations; new multiform synthesis units and a ne w speech modification algorithm based on a vocoder that offers harmonics re construction. The multiform units make it possible to reduce acoustic disco ntinuities at concatenation points and unnatural sound by preparing synthes is units with various lengths and various F-0 contours. The new speech modi fication algorithm, on the other hand, improves the quality of prosody modi fied speech. This algorithm is extremely effective in synthesizing speech w hose prosodic parameters are quite different from those of synthesis units. Listening tests confirm that the new synthesis units yield speech with hig h intelligibility and naturalness, and that the new speech modification alg orithm is superior to all other conventional vocoders and waveform domain a lgorithms including TD-PSOLA, especially when modifying the F-0 frequency u pward.