摘要 |
PROBLEM TO BE SOLVED: To perform efficient encoding for an acoustic signal including a voice and a singing voice of a human. SOLUTION: A vocal acoustic signal to be encoded is PCM(pulse code modulation)-encoded and taken in as acoustic data, and plural unit sections are set on a time base. Fourier conversion is performed for each unit section and a spectrum S is obtained. The prescribed threshold value L is set, a series of continuous part of which the intensity is this threshold value L or more out of the spectrum S is recognized as formant F1-F5 being peculiar in vocal respectively. The maximum peak frequency in each formant is extracted as a representative frequency representing the formant, and one representative frequency per formant is decided. Each representative frequency is replaced by a note number of MIDI(musical instrument digital interface) data, an acoustic signal in the unit section is encoded by this note number. |