LAKE-BOTTOM RECOGNITION USING A WIDE-BAND SONAR SYSTEM AND TIME-FREQUENCY ANALYSIS

The work presented in this paper deals with the response of underwater bottom insonified with wideband ultrasonic signals. The aim of the study is to investigate the feasibility of the recognition of the nature of sediment layer using a wideband sonar system and time-frequency representations of backscattered echoes from the bottoms. The experiments have been performed in a natural environment, in Lake Geneva. Time-frequency representations of backscattered echoes based on the pseudo-Wigner distribution are realized. A method of feature extraction using the energy distribution of the time-frequency images associated with the echoes is presented. Two classification techniques have been applied to recognize the nature of the lake bottom: a classical discriminant analysis and a neural network approach using multilayer perceptron architecture. The average recognition rates for five different types of lake bottom and for both techniques are presented and compared. © 1995, Acoustical Society of America. All rights reserved.