A new approach to frequency line tracking

The Dynamic Undersea Detection Extractor (DUDE), an implemented acoustic feature extractor designed to extract and statistically characterize narrowband signals in noise, is described. DUDE accepts normalized spectral data and produces symbolic representations of narrowband acoustic signals. If the data are beamformed, DUDE also performs multibeam correlation to produce bearing estimates. DUDE distinguishes itself from other feature extractors primarily by its ability to characterize signals that move in frequency over time (i.e., dynamic signals). Among the unique features in DUDE´s algorithm is the `line-likeness in a neighborhood´ test used to prune spurious signals. Statistical performance analysis results for the DUDE algorithm are also presented