基于RSSI概率统计分布的室内定位方法

Method for indoor positioning based on RSSI statistical probability distribution

针对接收信号强度指示(Received Signal Strength Indication,RSSI)时变现象影响WLAN室内定位精度问题进行了研究,提出了一种基于RSSI概率统计分布(Statistical Probability Distribution,SPD)的加权K最近邻(Weighted K-Nearest Neighbor,WKNN)方法——SPD-WKNN方法。该方法首先利用SPD方法得到指纹点RSSI向量区间;然后运用SVM算法选取测试点K个近邻指纹点,计算测试点RSSI向量到每个近邻指纹点的最小欧氏距离;最后结合WKNN算法获取定位结果。实验结果表明,SPD-WKNN方法与NN、KNN、WKNN、SVR和LSSVM方法相比定位误差分别降低了47.3%、41.6%、31.9%、27.1%和16.3%,呈现了良好的定位效果;利用SVM算法的稀疏性明显减小了运算时间。

The phenomena time-varying of Received Signal Strength Indication（RSSI）affects the indoor positioning accuracy in Wireless Local Area Network（WLAN）. A new Weighted K-Nearest Neighbor（WKNN）indoor positioning method based on Statistical Probability Distribution（SPD）（SPD-WKNN method）, is proposed to resolve the problem.Firstly, it gets the vector interval estimation of RSSI in Fingerprint Points（FPs）by SPD method. Then, it selects K nearest neighbor FPs by Support Vector Machine（SVM）algorithm and calculates the minimum Euclidean distance from the RSSI vector of test point to each nearest neighbor FP＇s vector interval of RSSI. Finally, it gets the positioning results by WKNN algorithm. The experimental results show that the proposed SPD-WKNN method reduces the average positioning error about 47.3%（1.47 m）、41.6%（1.17 m）、31.9%（0.77 m）、27.1%（0.61 m）and 16.3%（0.32 m）compared to NN、KNN、WKNN、SVR and LSSVM respectively. The operation time is obviously reduced by the sparsity of SVM classification algorithm.