Estimating Doppler shift with a coarse cochlear filterbank

Abstract

In this work, RoBat, a biomimetic sonarhead mounted on a mobile robot is presented as a platform in which to study perception in bats. After addressing the relation between bats and robots, a biologically plausible model of the bat's cochlea is described. The model consists of a filterbank with an artificial neural network (ANN) connected at the end of it. The ANN is used for extracting the frequency of the received echoes. Thus, Doppler shifts, which play a fundamental role in bat navigation, can be successfully estimated. Experiments with simulated and real data show the efficiency of the system when used on a LEGO mobile robot.