next up previous
次へ: Model Parameter Estimation using 上へ: A Maximum Likelihood Formulation 戻る: A Maximum Likelihood Formulation

Model of Harmonic Structures

An influence of a window function and a varying pitch within the short time single analysis frame inevitably cause widening of the spectral harmonics which makes it difficult to extract the precise value of s and to separate close partials. First we assume that each widened partial is a probability distribution of frequencies, approximated by a Gaussian distribution model. Therefore, a single harmonic structure can then be modeled by a constrained Gassian mixture model (a harmonic-GMM), in which their means have only 1 degree of freedom. In log-frequency scale, means of harmonic-GMM are denoted here as where ideally corresponds to the of th sound and denotes the index of partials. We then introduce a model of multiple harmonic structures which is a mixture of harmonic-GMMs whose model parameter is denoted as

    (1)



where and indicate the weights and variance (which is briefly assumed here as a constant) of the respective Gaussian distributions.
next up previous
次へ: Model Parameter Estimation using 上へ: A Maximum Likelihood Formulation 戻る: A Maximum Likelihood Formulation
平成16年3月25日