Localization of off-the-shelf mobile devices using audible sound

Sound source localization will play a major role in the new location-aware applications envisioned in Ubiquitous Computing. We describe the design and performance of three architectures and corresponding protocols that use a variation of the Time-of-Flight method for localizing three different kinds of devices, namely 802.11-enabled PDAs, 3G cell phones, and PDAs without network connectivity. The quantitative assessment is based on the deployment made with 6 sensors in, a 20x9m room, sewing over 10,000 localization requests. Our experiments indicate that all architectures achieve localization within 70cm of the actual position 90% of the time. The accuracy is further improved to 40cm 90% of the time when geometric factors are taken into consideration. The effects of noise and obstructions are also analyzed. Within 1m localization error realistic noise degrades the accuracy by 6 to 10%. The presence of obstacles, such as humans and cement columns, has no observable effect on the performance.