摘要 |
PURPOSE:To achieve a sound synthesization with limited noise avoiding extreme change in the frequency by memorizing frequency extracted from the main spectrum into a reference frequency memory provided in the main spectrum extractor to set the subsequent extraction frequency. CONSTITUTION:Frequency spectrum of input voice from an analyzer is stored into a spectrum envelope memory 201 of the main spectrum extractor 200'. Peak values at the crest of the spectrum envelope in the memory content are read out with a secondary peak level detection circuit 202 by N pieces sequentially from the maximum and hence, a frequency extraction circuit extracts the frequency at the peak point. The results of the extraction are fed to a synthesizer together with the results of the extraction with an amplitude extraction circuit 203, a frequency extraction circuit 204 and a phase extraction circuit 205 while being stored into a reference frequency memory 210 to be the reference for setting frequency through the memory 201 with a side peak detection circuit 211 when the subsequent peak point is extracted. Thus, main spectrum is extracted without extreme change in the frequency thereby accomplishing a voice synthesization based on a voice analysis with limited noise. |