摘要 |
PROBLEM TO BE SOLVED: To improve sound quality by preventing degradation of sound quality caused by pitch variation. SOLUTION: Sound sources g1 , g2 , g3 are individually inputted to acoustic pipes s1 , s2 , s3 , a time series waveform of one pitch section is extracted from sound sources and an acoustic pipe shape, that is, element data is extracted from the acoustic pipes s1 , s2 , s3 . Element data extracted from the acoustic pipes s1 , s2 , s3 are fast fourier-transformed by fast Fourier transformation sections FFT1 , FFT2 , FFT3 , and spectrum patterns are extracted to an output. They are spectrum-processed by spectrum processing sections SP1 , SP2 , SP3 using this spectrum pattern and values obtained from second and third Formant frequencies. After the spectrum-processed data converted to new element data by inversion fast Fourier transformation, the element data is synthesized by the synthesizing section, and synthesized voice is obtained. |