![]() | Only 14 pages are availabe for public view |
Abstract The thesis provides the analysis of acoustic echo fonnation and the basic properties of echo structure, then it describes a neural network system that is capable of recognizing different types of underwater acoustic signals, obtained from the back scattering of the underwater pulse modulated signals. The used sonar signals recognition system deals with the envelopes of the echoes received from complex geometrical shapes underwater targets instead of the echoes. The sonar signals are pre-amplified, filtered through the active sonar equipment, then the envelopes of these signals are detected. The envelopes are taken to extract the feature vectors that can characterize the targets by using two methods:- 1- Discrete Cosine Transfonn method (DCT) 2-Linear Prediction method (LP). After the processing of feature extraction , the application of the neural network as a classifier are presented by applying the back propagation algorithm using specific neural and the main conclusions extracted. From the main conclusions of display that the suggested methods can perfonn successfully the a- Recognition of complex geometrical shapes underwater targets. b- Simplifying the neural network used as a classifier. c- Decreasing the time required for learning and testing d- Increasing the efficiency of the recognition system |