摘要 |
PURPOSE:To reduce the constitution of a parameter memory to be used without deteriorating the quality of a synthesized voice by reading out a voice feature parameter in a two-sound component continuing part stored in a table memory in the case of a component name which is not stored in a component name memory and combining respective voice feature parameters to synthesize continuous voice. CONSTITUTION:A two-sound element table 6 expressing relation between two- sound component continuing type short component names and the storing positions of short component spectral parameters indicates the storage positions of the short sound component parameters in order to use a CVC (consonant/ vowel component) name table 3, and CVCs not stored in a CVC parameter memory 5 and having small generation frequency by synthesizing them from CV or VC short sound components. In this case, short component spectral parameters which are not stored in an exclusive memory are used by partially segmenting CVCs stored in the memory 5. Thus, a memory for storing the spectral parameters of two-sound component units can be saved. |