Interference detection using time-frequency binary hypothesis testing

The growth in demand for wireless broadband is increasing the instrumental value of characterizing interference in spectral bands of interest. In particular, detection and classification of interference is an important function for mobile ad hoc network (MANET) nodes to optimize performance. In this paper, we propose interference detection and classification techniques for a network that consists of N distributed nodes and a classification center connected to each node by a robust link. Due to the limited capacity of the network connecting these nodes, each node compresses the observed signal into a binary vector that spans time and frequency and provides the bits to a classification center. In turn, the classification center stitches the compressed signals from all nodes into a single structure with characteristics that are exploited to detect and classify. The Kolmorgorov-Smirnov test is proposed for node compression. Furthermore, bounds on theoretical classification performance from finite blocklength coding theory are defined. Finally, we propose a minimum distance method to perform classification.