Matching pursuit for the recognition of power quality disturbances

Signal decomposition and features extraction are essential and necessary for the study and the classification of different signal types. In this paper, the authors propose an appropriate approach in processing different types of power quality disturbances using a so-called adaptive signal representation tool called matching pursuit. Given a redundant dictionary of waveforms, they decompose a signal into a linear expansion of these waveforms, which are selected in order to best match the signal structure. In particular, using a dictionary of Gabor functions g_γ{γ=(s, u, ξ)}, they decompose a given disturbance into dominant atoms in the frequency-time distribution which characterise the type of power quality disturbance. This compact set of atoms, in conjunction with a radial basis function (RBF) classifier, is found to be very efficient for the classification of nonstationary and transitory power quality disturbances