The Information Theoretic Approach to Signal Anomaly Detection for Cognitive Radio
Figure 8
Analysis of a WLAN signal with interference from a Bluetooth device. (a) Spectrogram of the original signal. Data bursts from the Bluetooth device are clearly visible as high power, lightly shaded patches. (b) Information content of events (power density at any given time-frequency point) in the signal. (c) Anomalies detected (light patches) using a threshold of (7.15 bits) and a cluster length of 1. The result is noisy and there are a lot of false positives (appearing as singular, lightly shaded spots). (d) Anomalies detected when the threshold is left unchanged at and the cluster length is increased to 3. The outcome is now much cleaner with virtually zero false positives.
