摘要
针对位置指纹定位算法在训练阶段信号数据采集量大和定位精度不高的问题,提出一种压缩感知(CS,Compressed Sensing)与K均值改进支持向量机(SVM,Support Vector Machine)相结合的定位算法模型(CS-KSVM)。CS算法在训练阶段利用已采集到的部分参考点wifi信号强度数据对整个指纹信号库进行重构以降低信号采集工作量,再用K均值改进SVM算法来实现测试点的准确分类。实验仿真结果表明,CS-KSVM算法在相同采样点条件下的定位精度明显要高于传统定位算法,同时在相同定位精度条件下大大减少了定位需要的采样点数。CS-KSVM算法在3米之内的定位准确度可以达到93.2%。
In view of the problem that in the training phase location fingerprinting localization algorithm has large amount of signal data acquisition number and its positioning accuracy is not high, put forward a kind of local-ization algorithm model（CS-KSVM） based on compressed sensing （CS, Compressed Sensing） and the improved sup-port vector machine with K- means clustering （SVM, Support Vector Machine）. The CS algorithm in training stage refactors the whole fingerprint library using the collected some reference point wifi signals strength data in order to reduce the workload of signal acquisition, and then uses k-means to improve the SVM algorithm for the sake of achieving accurate classification test point. The experimental simulation results show that the positioning accuracy of CS-KSVM algorithm is significantly higher than that of traditional localization algorithm under the condition of the same sampling point, at the same time the sampling points of positioning are greatly reduced under the condition of same positioning accuracy. The positioning accuracy of CS-KSVM algorithm within 3 meters can reach 93.2%.
出处
《激光杂志》
CAS
CSCD
北大核心
2014年第9期82-85,共4页
Laser Journal
关键词
位置指纹
压缩感知
支持向量机
室内定位
wifi
Position fingerprint
Compressed sensing
Support vector machine
WiFi
Indoor positioning
