H. Satar-boroujeni and B. Shafai (USA)
Modeling, system theory, music analysis, tracking.
In this paper a state-space model for the evolution of frequency and power partials in music signals is presented. The noisy observations are power and frequency of spectral peaks from discrete time frames, which are detected using a novel technique. Our state space model is instrument-specific in the sense that for different classes of melodic instruments different models are introduced. Parameters of each model can be estimated through an extensive statistical analysis of a large number of musical sounds in a forward-problem procedure. These parameters are frequency-dependant and for each frequency bin a different set of parameters are estimated.
Important Links:
Go Back