A computational model of echo location

Abstract

The range and velocity of a moving object were measured unambiguously by a computational model of echo location. Based on the echo-location sounds utilized by some species of FM bats, a linear-period-modulated (LPM) signal was adopted in this model. The LPM signal was designed to yield a phase rotation of π rad for velocity of 0.5 m/s, and its frequency was hyperbolically swept down from 60 to 40 kHz within a 1.4-ms duration time. Experiments were performed in air by changing the velocity of a plastic disk of about 20 mm in diameter. When a passage of the disk was detected by a photosensor installed at a distance of 1 m from transducers, the LMP signal was emitted as a location sound. Echoes were processed by a bank of constant-Q filters, and the range and velocity were estimated according to the temporal phase distribution of phase-sensitive elements connected to the output of each filter. Regardless of the relatively small time–bandwidth product of the LPM signal, results of these estimations show that the estimation accuracy was better than 0.1 m/s in velocity and about 0.5 mm in range. On the contrary, a conventional sonar system with an up-chirp signal did not yield reliable results because of its ambiguous property in simultaneous parameter estimation.