Extraction of melody underlying the audio signal

摘要

The main melody, and is a part of the musical work to perceive people is the largest, to be the most accurate, in the case of assumptions careful consideration, it was more stable, the more apparent the automatic transcription or melody extraction, applicable If, costs are to take less. For this, in the judgment of the melody of the voice signal, First, I ask the melody line extending in the time / spectral representation. Or frame for each time segment, precisely, one frequency bin or one spectral component of the time / spectral representation, are uniquely associated. That is, by a special embodiment, the sound result of the maximum intensity is derived in this frame. [Selection Figure Figure 3