Inaudible Sounds From Appliances as Anchors: A New Signal of Opportunity for Indoor Localization

Abstract

The indoor localization still has challenges when compromising accuracy, cost, scalability, and so on. In recent years, signals of opportunity have been widely used for low-cost and scalable indoor localization services, especially in complex and changeable environments. In this article, a new opportunistic signal for indoor localization is proposed using the inaudible sound pattern from the appliances, which has superiority in long-term existence, stability, high availability, and privacy protection. The short-time Fourier transformed feature is exploited with a random forest model to recognize the appliance, and a promised matching algorithm is proposed to avoid any false detection. The recognized appliances can be used as standalone location indicators or supplements, such as loop closure pairs and additional landmarks, to the existed indoor localization system. The experimental results show that the appliance recognition achieves a recall rate of 79.66% under 100% precision, and a 95.43% improvement of localization accuracy is obtained to the pedestrian dead reckoning system over a 621.58-m trajectory.