| Musical signal processing is an important research area in signal analysis. This dissertation is concerned with the problem of automatically analyzing music signals including time frequency analysis, note identification and signal decomposition. Although this approach may be applied on signal created by other musical instruments, currently it is specifically designed for violin music.;The proposed system is a multiple step iterative approach inspired by the way human beings analyze music signals. Firstly, using a method named Instantaneous Fourier Decomposition, the signal is analyzed to get a form of time frequency representation result. Instantaneous Fourier Decomposition is a bandages notch filter bank based on Internal Model Principle. The second step is note extraction based on this time frequency representation with a Computational Auditory Scene Analysis method. At the end, with the note extraction result, the music signal is decomposed to Narbada components with an adaptive version of Instantaneous Fourier Decomposition.;Results on synthesized signal, MIDI and real recording are presented and discussed. |
