Syntactical classification of extracted sequential spectral features adapted to priming selected interference cancelers.

摘要

Multiple classes of narrowband interference can be observed in a communications channel. Traditional adaptive filtering methods cannot filter robustly over the entire possible range of classes of narrowband interference. This dissertation arose as a systems dissertation, from real-world research, to intelligently classify and mitigate multiple classes of narrowband interference in a channel. Several areas are combined into a new and innovative overall systems application to classify narrowband interference to prime selected adaptive interference cancelers, resulting in faster convergence times. The front-end classification is composed of extracting spectral features over time into sequences, followed by syntactical classification of the interference type. Interference spectral features are extracted via image processing tools, from a spectrogram of the channel data transformed into an image. Interference classes are represented by syntactical augmented grammars. Sequences of interference features are then classified by maximum similarity classification. Interference extracted features and the corresponding class label, prime selected interference cancelers for (near-) optimal filtering, resulting in fast convergence.