Scale detection for transcription remarking zero outputs of comb filters

Abstract

In this paper, we propose a method of the automatic transcription system remarking zero outputs of comb filters. In the frequency domain, the comb filter has zero points at the integer multiple of the fundamental frequency. The fundamental frequency and harmonic frequencies of the musical tone can be eliminated by comb filters which are corresponding to the scales. In the method using comb filters which has already presented, 60 comb filters are required, and the processing time is too long to detect the scale of the 32th note. In the proposed method, we can estimate the scale of both monophony and chords using only seven comb filters and a few notch filters connected in cascade, and we can also detect the scale of the 32th note (62.5ms). We have done some computer simulations on transcribing the melody played on a piano. As a result, it is clear that the proposed method is useful to detect musical scale.