信道状态信息指纹定位算法性能评价方法研究

基于WiFi的定位技术大多使用接收信号强度,但该方法受多径和噪声干扰较大,精度有待提高。信道状态信息(channel state information, CSI)能够更加精细地描述信道状态,具有更强的稳定性。将CSI作为格点特征建立指纹定位数据库,利用该指纹库和在线测量数据,比较了多种定位算法在位置指纹法中的定位效果,并提出了评价KNN、wKNN和随机森林算法的一种评价依据和样本容量扩充方法,分析了三种方法随样本容量增加时定位时间和定位精度的稳定性,从包含定位精度在内的多种角度更加全面地评估了三种方法。结果表明,在以上三种定位算法中,随机森林算法的定位时间与定位精度的稳定性最好。

Performance Evaluation Method of Fingerprint Localization Algorithm Based on Channel State Information

Most location techniques based on WiFi use received signal strength. However, this method is subject to multipath and noise interference, and its accuracy needs to be improved. CSI (Channel State Information) can describe channel state more finely and has stronger stability. In this paper, CSI is used as a lattice feature to establish a fingerprint location database. Using this fingerprint library and online measurement data, we compared the positioning performance of different location algorithms. We proposed an index for evaluating KNN, wKNN and random forest algorithm and a sample capacity expansion method. Then we analyzed the stability of positioning time and positioning accuracy of three algorithms when the sample size increased. We evaluated these algorithms not only according to positioning accuracy, but also from other aspects. The results show that, in the above three localization algorithms, the stability of positioning time and positioning accuracy of random forest algorithm are the best.