基于压缩感知的室内定位系统的定位性能分析

随着Wi-Fi技术的普及，Wi-Fi室内定位技术也越来越受到关注。压缩感知(compressive sensing， CS)技术被提出应用于Wi-Fi室内定位，为了研究各类CS算法在室内定位系统中的定位性能，构建出一套基于CS算法的室内位置指纹定位系统。在离线阶段采集数据并构建指纹库，在在线定位阶段采用不同压缩感知算法比较各类算法的定位性能。实验结果表明，设备朝向包含多方向，参考点数据量越多时定位性能更优；CS的算法参数会影响定位性能；在设定的实验环境下，压缩感知中的分段弱正交匹配追踪(stage-wise weak orthogonal matching pursuit， SWOMP)算法的定位精度比K最近邻算法(k-nearest neighbor， KNN)优21.9%；在各类压缩感知算法中，正交匹配追踪(orthogonal matching pursuit， OMP)相较于其他CS算法表现较差，并且这种差距随参考点数据量的增多而愈加明显。

Localization performance analysis of indoor positioning system based on compressive sensing

With the popularity of Wi-Fi technology, the WiFi-based indoor positioning technology has fostered growing interests. The compressive sensing (CS) theory has been proposed for the WiFi-based indoor positioning. In order to study the localization performance of the system with various CS algorithms, we construct a CS-based indoor location fingerprinting positioning system. In the system, we collect data and build a fingerprint database firstly in the offline phase, and then in the online phase we use different compressive sensing algorithms to compare the localization performance of the system. Experiments show that: first of all, the larger the reference point data set collected from multiple directions contain, the better the performance of location becomes. Secondly, the algorithm parameters of CS have certain influences on the positioning accuracy. Thirdly, the stage-wise weak orthogonal matching pursuit (SWOMP) algorithm in compressive sensing has a better localization accuracy than the k-nearest neighbor(KNN) algorithm by 21.9% in the experimental environment of the paper. And lastly, among all kinds of compressive sensing algorithms, the orthogonal matching pursuit (OMP) is worse than other CS algorithms, and this difference becomes even more obvious with the growth of reference point data.