Precise Multimodal Localization with Smart Phones 教程.pdfVIP

Precise Multimodal Localization with Smart Phones E. Martin and R. Bajcsy EECS Dept., University of California, Berkeley emartin@ Abstract. In this paper we propose the integration of computer vision with accelerometry and the magnetometer and radios available in current state-of- the-art smart phones, in order to provide a precise localization solution feasible for indoor environments. In terms of accelerometry, we apply the wavelet transform to the signal of a single off-the-shelf accelerometer on the waist to obtain the velocity and stride length of the user. This allows us to link accelerometry and computer vision through the kinetic energy of the user. Additionally, our system leverages the capabilities of current state-of-the-art smart phones to integrate both offline and online phases of radio fingerprinting with WiFi, achieving an accuracy of 1.5 meters. We have also studied the possibilities offered by the cellular communications radio, with the intention to build a multimodal solution for localization, delivering an accuracy of up to 0.5 meters when all the information is combined with a Kalman filter. Keywords: Localization, wireless networks, multimode. 1 Introduction Context awareness is a key factor for multiple applications, and location is a fundamental parameter to consider. The research community is focusing on multimodal systems for localization, since the combination of different technologies increases the robustness and accuracy of the solution. Most of the recently proposed appli