基于核化K-means和SVM分类回归的Wi-Fi室内定位算法

摘要

针对目前指纹室内定位系统指纹库管理效率低、实时性差和定位精度低的问题,提出了一种新的基于核化K-means和SVM分类回归的无线定位算法.首先利用核化K-means算法将输入的预处理后的RSS(Received Signal Strength)信号进行无监督聚类,将聚类后的数据信息存入指纹特征数据库,然后通过SVM回归的机器学习算法对特征数据库的数据进行训练,得到一种最优的拟合位置函数的数学模型.并且采用粒子群算法对参数进行寻优,进行实验仿真.实验结果表明,该算法有效地提升了定位精度,优于KNN、WKNN、SVR等室内定位算法.%Current indoor localization system has a low real-time performance,low precision,and bad efficiency of the fingerprint library.In order to solve the problem,this paper proposed a new indoor positioning algorithm based on kernel K-means and SVM classification regression methods.The algorithm firstly employed preprocessed RSS signal to conduct unsupervised cluster and then saved the data into fingerprint database when the kernel K-means algorithm had finished.Secondly,SVM learning machine made use of the input sample data to train and generated a mathematical model of optimal fitting position function.Finally particle swarm optimization (pso) algorithm was used for parameters optimization and the simulation experiment was carried out.The results of experiments show that the proposed algorithm effectively enhances the accuracy and is better than that of K-means,KNN and other localization algorithms.