Application of Gabor analysis for detection, estimation and classification of underwater acoustic data

The problem of detection of transient signals emitted by underwater objects and classification of these objects based on characteristics of transient signals, is addressed in this paper. Even though the detection of transient signals using an energy detector is one of the simplest ways, it is not an optimum detector when some information about the signal is known. However, the performance of the energy detector can be improved by using a detector based on Gabor analysis, since Gabor representation is inherently localized and thus more suitable for transient analysis. In this paper, a description of such a detector is given. In addition, an estimator which estimates signal parameters that are suitable for classification is described. These estimated signal parameters are then used to classify a given signal set. The classifier based on "multiple discriminant analysis" that is described in this paper, is used for this purpose. Finally, the performances of the detector, the estimator and the classifier are verified using underwater acoustic data. The experimental details and the results are also discussed.