Musical pitch estimation and discrimination of musical instruments using comb filters for transcription

We propose a new method for the musical pitch estimation and the discrimination of different musical instruments. Most previous methods are based on the extraction of the pitch frequencies. But the extraction of the pitches is difficult since the musical sound has many harmonic components and these components are changed complicatedly depending on the tones and musical instruments. The principle of our method is the elimination of the pitch and its harmonic frequencies. This is performed simply by a comb filter (H(z)=1-z/sup -n/). We can estimate the pitches by detecting the zero outputs of the cascaded comb filters. Then, we separate polyphonic music into each monophonic one using comb filters and we can estimate musical instruments by frequency analysis.