Choosing coherent times of flight for improved acoustic sensor localization

Abstract

Acoustic sensor localization is able to provide mobile users with contextual information without requiring additional hardware components. The location of a mobile device indoors can be computed based on its estimated distances from loudspeakers whose positions and emitted signals are previously known. The spatial information can be inferred from estimated times of flight (TOFs) and the speed of sound. In real environments, however, reverberation and line-of-sight obstructions impose considerable challenges to TOF estimation, with potential to significantly degrade the localization performance. This paper proposes a volumetric-search method that guarantees the use of coherent TOF information by discarding erroneous TOFs. In a simulated realistic scenario, the proposed method achieves localization errors of a few tenths of centimeters.