摘要 |
PROBLEM TO BE SOLVED: To provide a voice synthesizing technology that is able to minimize error and able to hold the ascending order of LSP parameters.SOLUTION: Using LSP parameters composing a series of phoneme HMM states in which ω^(t)(i=1,2, and so on. p) is used as an LSP parameter and a difference LSP parameter d(t) is set such that d(t)=ω^(t)>0 and d(t)=ω^(t)-ω^(t)>0(2≤j≤p), a difference LSP parameter calculating section 5 repeatedly updates and calculates a difference LSP parameter dby a multiplication update system. Using the calculated difference LSP parameter d, an LSP parameter calculating section 6 calculates an LSP parameter ω^(t) defined by the formula given below. Based on the LSP parameter ω^(t), the voice synthesizing section 3 generates a voice corresponding to an input text. |