| 摘要 |
PROBLEM TO BE SOLVED: To provide a phoneme data processor or the like in which the capacity of data used to synthesize voice can be reduced while maintaining tone quality at a high level and which can make each data usable. SOLUTION: Phoneme piece data inputted from a phoneme inputting section 1 are Fourier-transformed by an FFT computing section 2. Based on the transformed result, a fundamental frequency detecting section 3 specifies a fundamental frequency. A band specifying section 4 determines how to divide the phoneme piece data and notifies an analyzer 5 of the determination. The analyzer 5 generates a plurality of band component waveform data that express the item, in which the waveforms expressed by the phoneme piece data are divided into a plurality of band components, in accordance with the notification. A gain adjusting section 6 lines up the amplitudes of the waveforms of these band components. A difference computing section 7 generates difference data which represent the difference between the waveforms of the components of the band including the fundamental frequency and the waveforms of the components of other bands. A DPCM computing section 8 stores the ones having smaller amplitudes of expressed waveforms among the coded ones of the band component waveform data and the difference data in a phoneme dictionary. |
