| 摘要 |
PROBLEM TO BE SOLVED: To attain a singing marking device capable of marking even a piece of music without melody. SOLUTION: A user singing data obtained by sampling singing voices sung by the user to karaoke accompaniment sound generated by reproducing an accompaniment data, and a model singing data read from a karaoke data memory 13 synchronized with the reproduction of the accompaniment data are divided into frames of the predetermined number of data, and a voice featured value MFCC of the model singing voice is extracted from the model singing data in the divided frames, and the voice featured value MFCC of the user singing voice is extracted from a user singing data, respectively. The singing is marked by deciding the propriety from a degree of similarity between the extracted voice featured value FMCC of the model singing voice and the extracted voice featured value MFCC of the user singing voice. That is, since it is marked whether or not the words of the piece of music are sung correctly, even a piece of music without a melody called "rap" can be marked for the singing. COPYRIGHT: (C)2008,JPO&INPIT
|
