HF-signal surveillance: signal detection, classification and parameter estimation

Communication signal surveillance in the crowded HF-band (3-30 MHz) is a difficult task. There is no normed frequency spacing in the HF-band. Due to the long range of HF-signals, the signals may spectrally overlap at the receiver, even if the transmitter intended to use an open channel. Some signals are intentionally hidden next to strong interferences. A multitude of digital and analog signalling formats are used. We propose a communication signal surveillance system for short-time stationary signals. The signal modulation amplitude shift keying, frequency shift keying, minimum shift keying, binary and quarternary phase shift keying, single side band analog transmission and unmodulated carrier are considered. Center frequency, bandwidth and the dwell time of the signals of interest are assumed to be unknown. The bandwidth of the signals is up to 5 kHz. Multirate filterbanks, morphological processing of the time-frequency plane, linear prediction, polynomial classification, which is an analog to neural networks, and AR-parameter models are used to detect, classify and estimate the parameters of the digital communication signals. Using the digitised receiver E1800A, the DSP board S56X and a SUN workstation, the real time signal surveillance system was configured and tested. In the paper we give the principles of the signal surveillance system and first test results