基于KDDA和SFLA-LSSVR算法的WLAN室内定位算法

发布时间：2018-04-26 04:29

本文选题：接收信号强度 + 无线局域网　；参考：《计算机研究与发展》2017年05期

【摘要】：针对接收信号强度(received signal strength,RSS)的时变性降低WLAN室内定位精度的问题,提出了一种基于核直接判别分析(kernel direct discriminant analysis,KDDA)和混洗蛙跳最小二乘支持向量回归机(SFLA-LSSVR)的定位算法,该算法通过核函数策略将采集的各接入点(access point,AP)的RSS信号映射到非线性领域,有效提取了非线性定位特征,重组定位信息,去除冗余定位特征和噪声;然后采用LSSVR算法构建指纹点定位特征数据与物理位置的映射关系模型,采用SFLA算法优化该关系模型的参数,并用该关系模型对测试点的位置进行回归预测.实验结果表明:提出算法在相同的采样次数下的定位精度明显优于WKNN,ANN,LSSVR算法,并且在相同的定位精度下,采样次数较大减少,是一种性能良好的WLAN室内定位算法.
[Abstract]:Aiming at the problem that the time variant of received signal intensity (signal) reduces the indoor positioning accuracy of WLAN, a localization algorithm based on kernel direct discriminant analysis (kernel direct discriminant analysis) and shuffled leapfrog support vector regression machine (SFLA-LSS) is proposed, which is based on kernel direct discriminant analysis (kernel direct discriminant analysis) and shuffled least-squares support vector regression machine (SFLA-LSSVR). The algorithm uses kernel function strategy to map the RSS signals of each access point to the nonlinear domain, effectively extracts the nonlinear location features, recombines the location information, and removes redundant location features and noise. Then the LSSVR algorithm is used to construct the mapping relationship model between the feature data of fingerprint location and the physical location. The parameters of the relationship model are optimized by SFLA algorithm and the location of the test points is predicted by the relationship model. The experimental results show that the proposed algorithm is superior to the WKNNNANN LSSVR algorithm under the same sampling times, and the sampling number is greatly reduced under the same positioning precision. It is a good WLAN indoor localization algorithm.
【作者单位】：合肥工业大学计算机与信息学院;芜湖创业园留学人员博士后科研工作站;
【基金】：国家科技支撑计划项目(2013BAH52F01)~~
【分类号】：TP18;TN925.93

【相似文献】