Biomimetic Sonar and Neuromorphic Processing Eliminate Reverberation Artifacts

Abstract

A biomimetic sonar is configured using two conventional sensors that generate point processes related to echo waveform intensity, resembling biological action potential spikes. The sensors point slightly outward from the sonar axis, similar to pinnae in some bats, to acquire slightly different views of the environment during a rotational scan. Artifacts in sonar maps are points that do not relate to actual object locations. Physical criteria identify artifacts by applying echo strength, azimuthal extent, and binaural coincidence criteria. Neuromorphic processing implements these criteria with thresholding, delays, and short-term memories. Artifacts are deleted to produce robust sonar maps. Multiple resolution maps, generated by using two thresholds, illustrate improvements over conventional sonar maps and tradeoffs between resolution and stability