基于FCM和DE-GPR的指纹库构建方法  

A fingerprint database construction method based on FCM and DE⁃GPR

作　　者：郑沛 张爱军[1] ZHENG Pei;ZHANG Aijun(School of Mechanical Engineering,Nanjing University of Science and Technology,Nanjing 210094,China)

机构地区：[1]南京理工大学机械工程学院,江苏南京210094

出　　处：《电子设计工程》2024年第23期51-56,共6页Electronic Design Engineering

摘　　要：针对室内定位中指纹库构建存在人力成本高、构建效率低的问题,提出一种基于模糊均值聚类(FCM)和差分进化算法(DE)优化高斯过程回归(GPR)的指纹库构建方法。FCM-DE-GPR算法通过模糊均值聚类和隶属度阈值划分指纹库局部模型,以挖掘无线信号的局部分布特征,并采用差分进化算法对GPR的超参数寻优过程进行改进,提高GPR的拟合精度。预测阶段根据K近邻原则确定扩展点的簇归属,使用相应的局部GPR模型进行预测,完成指纹库扩展。通过仿真与实验验证,所提算法将指纹库扩展至147%的情况下,平均预测误差相较于全局GPR降低了8.1%,具有良好的指纹库扩展精度。Aiming at the problems of high labor cost and low construction efficiency in fingerprint database construction in indoor location,a fingerprint database construction method based on Fuzzy C-Means(FCM)and Differential Evolution(DE)optimized Gaussian Process Regression(GPR)was proposed.FCM-DE-GPR algorithm divides the local model of fingerprint database through fuzzy mean clustering and membership threshold to mine the local distribution characteristics of wireless signals,and uses Differential Evolution to improve the super parameter optimization process of GPR to improve the fitting accuracy of GPR.In the prediction stage,the cluster ownership of the extension points is determined based on the K-nearest neighbor principle,and the corresponding local GPR model is used for prediction to complete the expansion of the fingerprint database.Through simulation and experimental verification,the proposed algorithm expands the fingerprint database to 147%,reducing the average prediction error by 8.1%compared to the global GPR,and has good fingerprint database expansion accuracy.

关 键 词：室内定位 位置指纹 模糊均值聚类 高斯过程回归 差分进化算法 

分 类 号：TN96[电子电信—信号与信息处理]