一种改进的基于KNN的动态预测指纹定位算法*

针对目前的WiFi室内定位普遍存在定位精度不高、定位结果不稳定等问题进行了研究,为了改善这些不稳定因素,根据室内传播信号波动较大的特点提出了一种基于KNN的指纹定位改进算法。该算法通过动态预测节点位置,从无线地图中过滤掉到标签处没有相似RSS向量的RP来寻找最近邻,以降低KNN算法的时间和计算复杂度。实验结果表明,改进后的算法在定位精确度方面有了较大的提高。因此,得出结论：改进后的KNN定位算法在减小位置漂移和定位的平均误差, 确实可以提高定位的精确度。

An improved Dynamic Prediction Fingerprint Localization Algorithm Based On KNN

Because there are some problems in WiFi indoor positioning system such as the low positioning accuracy and the instability positioning results, in order to improve these unstable factors, the paper deeply studied KNN fingerprint localization algorithm and improved the algorithm according to the characteristics of signal propagation volatile in the indoor environment. The algorithm finds the nearest neighbor through dynamically predicting node position and filtering out the RP without similarity RSS vector at labels from wireless map in order to reduce time and computational complexity of the algorithm KNN. The experimental results showed that the improved algorithm has been greatly improved in terms of location accuracy. Therefore, it is concluded that the improved KNN algorithm can improve the accuracy of positioning, which can reduce the mean error of position drift and positioning.