Classification And On-Line Learning Of The Acoustic Transient Signals

Underwater target classification is an important part of underwater acoustic engineering. With the development of ship-noise suppressing technology, the steady signals radiated by underwater targets are reduced, so it becomes more difficult to detect and recognize targets. Transients are difficult to restrain for they are radiated by the target randomly, so the research on the detection and recognition of transients is a new method of the underwater targets recognition. In this paper transients received by the passive sonar are analyzed and their features in both time domain and time-frequency domain are extracted for target classification. To solve the problem that the data set is non-exhaustive, a Fuzzy ARTMAP network is applied to design a transient signal classifier which has an on-line learning ability.The contents of the paper are arranged as follows:In chapter 1, the purpose of the paper is given and an overall classification system is designed. Several kinds of common transient signals are introduced in chapter 2. The structures and principles of Fuzzy ARTMAP network are presented in chapter 3. In chapter 4, according to the multibeam receiving characteristics of the sonar, a method of detecting transients in the beam space is proposed. In chapter 5, features in time domain and the time-frequency domain are studied. Several time-domain features and energy distributions extracted from real data are compared. In the last chapter, the performances of classification and on-line learning of Fuzzy ARTMAP with simulation data and real data are investigated, and the results show that the network can learn the new samples quickly, with a good adaptability.