| 摘要 |
PURPOSE:To output a synthesized voice which is close to human uttering by generating a parameter column employing the continuation time length of phonemes, generating a pitch pattern based on the phoneme continuation time length and phoneme information and synthesizing voice waveforms by the parameter column and the pitch pattern. CONSTITUTION:A phoneme continuation time length information is stored in advance in a dictionary 24 and the continuation time length of each phoneme is computed. The computed continuation time lengths are outputted to a parameter connection section 27 and a pitch pattern generating section 28. In the section 27, the phoneme piece data selected from a phoneme piece database 29 are held based on the phoneme information from a symbol column analysis section 25 and the continuation time length computed by a phoneme continuation time length computing section 26 and a parameter column is generated. On the other hand, a pitch pattern is generated in a pitch pattern generating section 28 based on the accent specification values from the phoneme information and the dictionary 24 and the continuation time length. In a voice synthesis section 30, synthesized voice signals are obtained based on the parameter columns and the pitch patterns. |
