Signal-tuned Gabor approach for power disturbance classification

摘要

The signal-tuned Gabor transforms are variant Gabor transforms based on temporal or spectral kernels whose parameters are defined, respectively, by the Fourier and inverse Fourier transforms of the analyzed signal. This yields time-frequency transforms which are tuned to the specific features of their inputs, thus allowing analysis with high conjoint spectro-temporal resolution. Here we investigate the use of the signaltuned Gabor transforms as tools for the classification of electric power waveform disturbances - namely, sags, swells, interruptions, harmonics, transients, spikes and notches -, following an approach based on the similarity of the time-frequency transform matrices. We show that the signal-tuned Gabor transform based on the spectral kernel yields higher-quality, more informative spectrograms, and reaches classification rates which are consistently superior to those afforded by the S transform.