A Fuzzy-Logic model for impulsive noise in PLC

In this paper an unsupervised classification procedure is proposed for the identification of different types of impulsive noise affecting an outdoor high voltage power line channel. Clustering of the time domain recorded data is preformed by a feature-based approach. Principal component analysis (PCA) is performed on the normalized spectra and clustering is performed on the principal components using a subtractive clustering method and fuzzy c-means (FCM). Several types of disturbances are recognized, arriving from different kinds of excitation. Finally coloring AR filters are designed for modeling each kind of noise. The proposed procedure has been applied to experimentally recorded noise data.