基于AP-SVM混合分类的指纹定位算法优化

Optimization of fingerprint localization algorithm based onAP-SVM hybrid classification

针对指纹定位技术在离线阶段采集数据样本时,因样本质量不稳定导致定位准确率不高的问题,提出基于亲和力传播-支持向量机(AP-SVM)结合黄金分割法的一个新的样本优化方法,即GAP-SVM,使定位准确率得到提高。首先进行接收信号强度指示(RSSI)值的采集并使用GP-AP聚类算法优化数据集,得到高质量、小样本的SVM分类器的训练集;然后设计出了基于GAP-SVM模型的位置指纹定位整体系统架构;最后进行实验仿真。实验结果表明:GAP-SVM混合分类器与传统的SVM分类器相比,GAP-SVM混合分类器具有更高的分类准确率,分类准确率达到92.21%,定位准确率比已有的指纹定位算法提高了46%。

Aiming at the problem that the location accuracy is not high due to the instability of sample quality when fingerprint location technology collects data samples in offline phase,a new sample optimization method,GAP-SVM,based on affinity propagation(AP)SVM combined with golden section method is proposed to improve the positioning accuracy.Firstly,the received signal strength indication(RSSI)value is collected,and the GP-AP clustering algorithm is used to optimize the data set to obtain the training set of SVM classifier with high quality and small samples.Then,the overall system architecture of position fingerprint positioning based on GAP-SVM model is designed.Finally,the simulation is carried out.The experimental results show that the GAP-SVM hybrid classifier has higher classification accuracy than the traditional SVM classifier,the classification precision reaches 92.21%,and the positioning precision is 46%higher than the existing fingerprint localizaion algorithm.