点击次数：

论文类型： 会议论文

发表时间： 2015-11-01

收录刊物： EI、Scopus

页面范围： 7-12

摘要： FM-based Device-Free Localization and Activity Recognition (FDFLAR) is a novel emerging technique which could sense location and activity information of a target utilizing only the ambient FM signals. FDFLAR realizes context aware with nearly no extra cost, which makes it a promising and attractive technique in future pervasive and ubiquitous computing applications. However, as a new technique, there are still lots of challenges to be solved. One fundamental problem is how to improve the accuracy of FDFLAR. In this paper, we explore methods to improve the accuracy of FDFLAR from two aspects. Specifically, on one hand we model FDFLAR as a sparse representation classification problem so as to improve the classification performance, on the other hand, we use joint frequency and space diversity scheme to improve the discernibility of FM features. Extensive experiments performed in a clutter indoor laboratory reveal the good performance of the proposed methods. ? 2015 ACM.