A sonar target recognition experiment

Abstract

This paper presents results of an experiment to determine the feasibility of an active sonar target recognition system. Echo data were obtained from 16 targets of maximum dimension 12 in., submerged in a salt water pool. The target frequency responses in the 15–45-kHz band were used to characterize the targets. The information in the frequency response was extracted in two ways: (1) by calculating the energy in a bank of filters, and (2) by modeling the frequency response as an all-pole filter. Pattern recognition techniques were applied to define and test the two recognition methods. The results show that target recognition is feasible with either of the methods tested. One hundred percent correct recognition is achieved with 15 parameters when no noise is added to the test echoes. If white noise is added to the test echoes, accuracy is well above 90%, provided the signal-to-noise ratio is at least 5 dB.