Non-stationary signal classification via modified fuzzy C-means algorithm and improved bacterial foraging algorithm

In this paper, empirical mode decomposition is applied to decompose the non-stationary power signals that results in a set of maximum and minimum points while satisfying the properties of the sifting process. Further, the empirical mode decomposition method is implemented to extract various intrinsic mode functions from the non-stationary signal disturbance waveforms that are already superimposed by various undulating modes. A novel short-time modified Hilbert transform with an equivalent window is applied on all the intrinsic mode functions to extract the modified Hilbert energy spectrum and instantaneous magnitude response. Distinct features are derived from the short-time modified Hilbert energy spectrum for automatic classification of non-stationary power signals. The features obtained from the short-time modified Hilbert transform are found to be different, understandable, and immune to noise. These features are then applied to the modified fuzzy C-means based improved bacterial foraging optimization algorithm for improving the classification accuracy of the disturbances. Extensive simulation results yield excellent visual detection, localization, and classification of the different types of non-stationary power signal disturbances. Copyright (C) 2016 John Wiley & Sons, Ltd.