Abstract: The indoor positioning technology based on fingerprint attracts the attentions from many researchers. RFID（RadioFrequency Identification） technology is more attractive, due to its high accuracy and adaptability to different environment. However, because of the heavy consumption of the computing power, it limits the applications in practice. A new hybrid Kmeans and Weighted K-nearest Neighbor method is proposed and applied in real-world indoor positioning. The new method divided the mapping area into several classes based on a clustering method. A matching into class is done first and then location is determined. The result shows that the proposed method reduces the accumulated errors and thus reduces the computational power whist maintains reasonable accuracy.

Key words: Radio Frequency Identification（RFID）, fingerprinting, indoor_positioning, kmeans, Weighted k-Nearest Neighbor

摘要： 基于指纹的RFID室内定位技术，由于其定位精度高、普适性强等优点受到国内外学者的广泛关注。但因为其计算量较大，在实际应用仍然非常有限。提出基于实际应用场景的Kmeans和Weighted K-Nearest Neighbor（WKNN）联合的定位方法，将指纹地图通过聚类算法分成块，先初步确认待测点所属指纹块，在块的基础上定位，这样可以减小误差累计。仿真结果表明，该方法在保证适当定位精度的同时，也减少了计算量和在线定位时间。

关键词: 无线射频定位技术（RFID）, 指纹, 室内定位, kmeans, 加权K近邻