Feature extraction for HF-band emitter location

A procedure for automatic feature extraction from emitters detected in wideband spectra is presented. Detection is based only on the analysis of the bearing angle spectrum with methods of digital image processing. Two different algorithms realize the detection of frequency stationary and also frequency agile signals. By combining both algorithms a pre-classification of all emitters into five classes can be achieved. The emitters are described by symbolic information like parameters in the time frequency domain and by an estimated bearing angle. Additionally, a spot check from the power spectrum is obtained which is related to a selected bearing angle range. The spot check as subsymbolic information is the basis for further methods of bearing angle verification and an advanced emitter classification. The procedure presented also yields an approach for data fusion between power and bearing angle spectra. The discussion of the results points out the data reduction potential of feature extraction which is essential in designing a wideband emitter location system.