| 摘要 |
<P>PROBLEM TO BE SOLVED: To synthesize a voice of high naturality and high continuity which accords with human speaking motions. <P>SOLUTION: A voice synthesizing method has a vocal tube filter determining resonance characteristics of a sound source and a voice. When control parameters for controlling the sound source and the vocal tube filter are used to synthesize a voice corresponding to input data (a phonemic sequence and the duration length of each phoneme), physiological measurement data of input data (pertinent phonemes) is retrieved from physiological measurement data preliminarily accumulated by simultaneously measuring vocal tube opening/closing movements and vocal cord opening/closing movements resulting from movements of jaws, lips, and a tongue for human speaking, and physiological measurement data is statistically analyzed in advance. Further, times of feature points on movement trajectories of the vocal tube and vocal cords are expressed as linear functions of duration lengths of phonemes to extract a control rule which determines a time relation between movement trajectories of the vocal tube and vocal cords, and movement trajectories of the vocal tube and the vocal cords corresponding to duration lengths of phonemes are created on the basis of the control rule to determine control parameters of the sound source and the vocal tube filter. <P>COPYRIGHT: (C)2004,JPO&NCIPI |
