基于近邻传播算法的动态自适应室内指纹定位算法

目前传统的室内指纹定位算法中存在以下几个问题，首先在构建指纹库时采用平均值的方式构造指纹库容易受到噪声点影响而降低定位精度，其次使用欧式距离衡量待定位点与指纹点之间的距离可能引入信号强度距离较近，物理距离较远的参考点参与估计待定位点的位置从而增大定位误差，以及当参考点数量较大时，由于K近邻算法的计算量较大，从而造成定位过程耗时较大，能源耗费较多的情况，除此之外，由于K近邻算法无法根据实际情况确定参与定位的参考点个数从而限制了定位系统的精确性和拓展性。针对上述问题，本文设计了一种基于近邻传播算法的动态自适应室内指纹定位算法。该算法在离线阶段对在每一个参考点采集的信号强度值使用方差滤波算法去除噪声值，然后利用加入了参考点物理信息的近邻传播算法对参考点进行聚类处理。在在线阶段，通过进行粗略定位和精确定位动态的估计待定位点的物理位置。经过实验证明，本文所提出的新算法较对比算法有较高的精确度和稳定度。

Dynamic and adaptive fingerprint indoor location algorithm Based on affinity propagation algorithm

There are several problems of the conventional indoor fingerprint localization algorithm currently. First construcing the fingerprint database using the average of data is susceptible to noise points and thus reduce the effects of positioning accuracy. Secondly using Euclidean distance to measure the distance between the point to be positioned and the fingerprint point may be introduced reference points which have the short distance of the fingerprint signal strength, but farther physical distance to be involved to estimated position of the point to be positioned thereby increasing the positioning error. When the number of reference points is large, since the amount of calculation of K nearest neighbor algorithm is large, the positioning process takes more time and more energy. In addition, since the K nearest neighbor algorithm is unable to determine the number of reference points which are used to locate the point to be positioned according to the actual situation, thus positioning system accuracy and scalability are limited. In response to these problems, we designed an dynamic and adaptive indoor fingerprint localization algorithm based on affinity propagation algorithm. The algorithm uses the variance filter to remove noise of the signal strength values collected at each reference point, and then cluster reference points using affinity propagation algorithm which combines the physical information of reference points in offline stage. In the online phase, dynamic estimate the physical location of the point to be located by perfoming coarse positioning and precise positioning. Experimental results show that the new proposed algorithm has a higher accuracy and stability than comparison algorithms.