Ultrasonic robot eyes using neural networks

Abstract

In order to construct a robot eye using ultrasonic waves in air, a system that combines existing acoustical holography with neural networks is devised. In this system, neural networks are used to reduce the vagueness of reconstruction from the acoustical holography. The system is able to identify objects and to reconstruct the acoustical images of objects using an array of only a small number of receivers. Some experimental results on identification and reconstruction using several kinds of metal plate objects are reported.