基于偏度-峰度检验的WLAN位置指纹室内定位算法

目前，无线局域网（Wireless Local Area Networks，简称WLAN）技术因其成本低、配置简单、精度高等特点，被认为是室内定位的最佳选择之一。虽然WLAN接收信号强度指标(Received Signal Strength Indicator，简称RSSI)指纹法是最精确的定位方法，但由于其无线电地图(Radio Map，简称RM)在发生环境变化时已经过时，具有很大的缺陷，且重新校准RM是一个耗时的过程。因此，本文提出基于偏度-峰度检验进行WLAN位置指纹室内定位算法改进。在离线阶段，通过偏度-峰度检验样本总体是否服从高斯分布，对于严重偏离高斯分布的样本直接舍去，而对于与高斯分布接近的样本，利用核函数估计其概率密度。在线阶段，利用K最近邻(K-Nearest Neighbor，简称KNN），将移动终端设备接收到的RSSI与建立的指纹数据库中的RSSI，通过欧几里得公式计算样本点到观测点的欧氏距离，并从中选择欧氏距离最短的样本点的位置作为研究位置的无偏估计。通过实验结果，本文提出的算法比传统定位算法的精度提高了11%，证明了该算法具有更高的定位精度和更少的离线工作量等优点，而且在RSS(Received Signal Strength)信号容量较小时该算法的定位误差比其他算法更小，具有显著的稳定性。

Indoor Localization Algorithm of WLAN Position Fingerprint Based on Skewness-Kurtosis Test

Nowadays, Wireless Local Area Networks (WLAN) is considered as one of the best choices for indoor localization because of its low cost, simple configuration and high precision.Although WLAN received signal strength indicator (RSSI) fingerprint is the most accurate positioning method, but its radio map (RM) is out of date when its environment changes, so it has serious defects. In addition, to recalibrate RM is a time-consuming process. In this paper, an indoor localization algorithm based on skewness-kurtosis test is proposed.In the off-line phase, the skewness-kurtosis test method is used to test whether the sample population is satisfied with the Gaussian distribution, the samples that deviate seriously from the Gaussian distribution are directly given up, and for samples that are close to the Gaussian distribution, the probability density is estimated by using the kernel function. In the online phase, algorithm of k-Nearest Neighbor is used to calculate the Euclidean distance from the sample point to the observation point by using the Euclidean formula between the RSSI received by the mobile terminal device and the RSSI in the established fingerprint database. Then the position of the sample point with the shortest Euclidean distance is selected as the unbiased estimation of the study position. Experimental results show that the precision of algorithm based on skewness-kurtosis test is 11% higher than the traditional localization algorithm. It is proved that this algorithm has the advantages of higher positioning accuracy and less off-line workload, and the positioning error of this algorithm is smaller than other algorithms when RSS signal capacity is small, even that it has remarkable stability.