The musical spectrogram : a tool for the time-frequency analysis of musical signals
Date
2002
Authors
Antoniou, Anthony
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The frequency domain of signals is traditionally examined through techniques based on the Fourier transform: These techniques give a linear view of the frequency spectrum in Hertz, and are ideal for the numerical analysis and manipulation of spectra. In this work, we propose a scheme suited to the analysis of musical signals: The Musical Spectrogram. A four-stage system, the Musical Spectrogram consists of a tuning module, a 9 octave band filter bank, a 12 semitone-band filter bank array and an output system. The spectrum of a musical signal is first decomposed into octave subbands, and each is separated into musical scale semitone subbands and plotted. New insights are offered by exploring the musical spectrograms of a number of test signals, and the commercial potentials of this new analysis tool are explored.
Simulation results of a Musical Spectrogram system are presented. A complete MAT LAB implementation of this system is tested to prove the function and form of the system. A limited real-time implementation of this system on the Analog Devices ADSP-2181 fixed-point digital signal processor is also presented which underscores the many difficulties inherent in development for such platforms, and details several novel solutions devised to address them. The numerous advantages and properties of the Musical Spectrogram system are also presented and discussed, and a complete real-time software version is put forward as future work.