The Anomaly Mixed Spectrum Signals Detection Based on ICA and KNN

Abstract

Independent Component Analysis is very effective for blind source separation of mixed signals, except detecting abnormal signal, presented an anomaly mixed spectrum detection method based on K Nearest Neighbor classification, using Independent Component Analysis to separate the mixed signal under Gaussian noise, and classifying the mixed signal based on training sample set, so as to find the abnormal signal mixed in the original signal. Simulation results showed that this method could not only separated signals effectively, but also improved the classification accuracy.