| 摘要 |
PURPOSE:To stably obtain the best approximation for the pitch pattern of a natural speech and to obtain a natural synthesized speech which is free from the feeling of physical disorder. CONSTITUTION:According to parameters which approximate the straight lines of the rise part and fall part of an accent component and are calculated by an accent parameter calculation part 5, an accent component generation part 6 interpolates the flat part of the accent component with a straight line to generate the pattern of the accent component. According to a parameter regarding a voicing component which is calculated by a voicing parameter calculation part 8, a voicing component generation part 9 generates the pattern of the voicing component. The generated patterns of the accent component and voicing component are added by an addition part 10 to generate a pitch pattern. |
