智能手机室内背景声感知与定位  

Indoor Background Sound Perception and Location Using Smartphone

作　　者：昂晨 王玫 罗丽燕 宋浠瑜 熊璐琦[1,2] ANG Chen1,2, WANG Mei1, LUO Liyan , SONG Xiyu1,2, XIONG Luqi1,2(1. Ministry of Education Key Laboratory of Cognitive Radio and Information Processing, Guilin 541004, China; 2. School of Information and Communication Engineering, Guilin University of Electronic Technology, Guilin 541004, China)

机构地区：[1]认知无线电与信息处理省部共建教育部重点实验室,广西桂林541004 [2]桂林电子科技大学信息与通信学院,广西桂林541004

出　　处：《无线电工程》2018年第8期650-654,共5页Radio Engineering

基　　金：国家自然科学基金资助项目(61771151);广西自然科学基金资助项目(2016GXNSFBA38014);中国博士后科学基金资助项目(2016M602921XB);广西高校无人机遥测重点实验室开放基金资助项目(WRJ2016KF01);广西研究生教育创新计划基金资助项目(YCSW2017137)

摘　　要：利用感知的背景声进行定位是室内定位的一种新型解决方案。拟在无需其他基础设施的条件下,利用智能手机麦克风被动感知室内背景声,实现房间级的定位。采用有监督的学习方式,对房间背景声进行功率第五百分位特征提取,并以此为训练集样本。分别使用k最近邻(KNN)、BP神经网络以及径向基神经网络(RBF)训练算法训练出房间背景声定位模型。以房间识别率与定位耗时为指标,通过线下训练、线上测试的方式对所提方案进行可靠性与有效性的验证。实验结果表明,RBF神经网络房间识别率在75%以上且运行时间较短。The positioning using the perceived background sound is a new solution for indoor positioning.This paper intends to use the smart phone microphone passive perception of the indoor background sound,room-level positioning without additional infrastructure.In this paper,the supervised learning mode is adopted to extract the fifth percentile power of the background of the room,which is used as a training set sample. The background sound localization model of the room is trained by k nearest neighbors（ KNN）,BP neural network and radial basis neural network（ RBF） training algorithm respectively.By using the room recognition rate and positioning time as an indicator,the reliability and availability of the proposed solution are verified through the offline training,online testing. The experimental results show that the recognition rate of RBF neural network is above 75% and the running time is short.

关 键 词：室内定位 智能手机 背景声 RBF神经网络 

分 类 号：TN915.04[电子电信—通信与信息系统]