METHOD FOR SYNTHESIZING SINGING, APPARATUS AND RECORDING MEDIUM

摘要

PROBLEM TO BE SOLVED: To realize synthesis of natural singing (B) which closely approximates human singing (A), on the basis of performance data inputted in real time.;SOLUTION: As performance data, there are inputted, for every phoneme such as 'sa', etc., which constitutes a lyric: state transition control information such as phoneme information, singing time information, singing length information, singing pitch information and attack; and effect control information such as vibrato, etc. The singing time information expresses the time for starting real singing such as T1, and each performance data are inputted earlier than the starting time of real singing. The phoneme transition time length is acquired from the database, on the basis of phoneme information for every performance data, and the singing start time and singing duration are determined for every phoneme, such as 's', on the basis of this phoneme transition time length, the singing start time information and singing length information. The singing start time of 's' can be set earlier than T1. In singing synthesis, singing voice is generated for every phoneme at the singing start time and the voice generation is continued over the singing duration. Transition state characteristics and effect characteristics may be controlled, according to the control information concerning the input.;COPYRIGHT: (C)2002,JPO