Robust Detection of Motor-Produced Audio Signals

Abstract

Indoor localization systems cannot rely on the same mechanisms, like GPS, that are used for outdoor or large-scale localization. Instead, autonomous or user-carried devices are often localized by measuring the time taken for an emitted signal to reach a known location; this signal can be sound, light, radio waves, or another similar sensed quantity. Autonomous mobile devices already contain motors, which produce sounds as a side effect of their operation, and so can potentially be included in a localization scheme without new hardware. In this paper, we briefly outline the challenges that need to be met for accurate detection and identification of motor-produced signals. We present a method for improving signal resolution for linear chirps that improves cross-correlation based signal detection by up to 2.8X.