室内基于声信号的智能移动终端非视距定位方法研究

发布时间：2018-07-01 11:54

本文选题：智能移动终端 + 室内定位　；参考：《浙江大学》2017年硕士论文

【摘要】：近年来,基于室内位置的服务需求变得越来越强烈。基于声信号的室内定位技术,由于具有兼容性好、稳定性和定位精度高,使其成为近几年的前沿研究课题。当前大多数的智能移动终端均具有扬声器和麦克风,这使得此类系统极易在实际环境中得到应用与推广。但实际上在真实的室内环境中,存在人员走动、家具遮挡、墙壁反射等等复杂的因素,使得声源与接收器之间的直接路径被遮挡,称为非视距(Non-Line-of-Sight,NLOS)。由于距离的量测依赖于声信号的时延估计,NLOS传播往往会给距离的估计带来较大误差,使得传统基于视距(Line-of-Sight,LOS)环境下提出的定位算法失效,导致定位精度急速下降。本文针对室内NLOS/LOS复杂环境,提出了一种基于NLOS判别的非视距定位方法,以及基于声信号信道统计特征NLOS识别的定位算法。本文的主要工作内容和贡献包含以下几个方面:首先,基于室内声音传播模型,对LOS环境和NLOS环境下声音信号信道特性进行研究。针对室内复杂环境中的强多径传输以及多普勒频移等因素,提出了基于互相关的声信道相对参数估计方法(相对时延及相对增益),降低了多普勒效应的影响,降低了计算复杂度。并提出一种基于信噪比(SNR)的自适应阈值函数,对声信号第一径到达时刻进行判别,降低了多径传输的影响。其次,基于声信道参数估计,对信道特性的特征提取进行研究,包括时延特性、波形形状特性、RicianK系数,以及增益的幅值分布特性,共9个用于NLOS识别的特征值,利用基于支持向量机的分类器,实现对NLOS信号的识别分类,并对其核函数及特征值组合进行了最优选取。最后,在NLOS信号识别的基础上,针对静态目标的NLOS定位,提出了基于NLOS识别剔除的定位策略和基于NLOS后验概率的加权最小二乘定位策略。针对移动目标提出了基于NLOS识别的修正卡尔曼滤波追踪算法和基于NLOS后验概率修正卡尔曼滤波追踪算法,以实现室内NLOS/LOS混合环境下的鲁棒定位追踪。
[Abstract]:In recent years, the service demand based on indoor location has become more and more intense. Because of its good compatibility, high stability and high positioning accuracy, acoustic signal based indoor positioning technology has become a frontier research topic in recent years. At present, most intelligent mobile terminals have loudspeakers and microphones, which makes such systems easy to be applied and popularized in real environment. But in the real indoor environment, there are complicated factors such as walking of personnel, furniture occlusion, wall reflection and so on, which make the direct path between the sound source and the receiver blocked, which is called Non-Line-of-SightNLOS (Non-Line-of-SightNLOS). Because the measurement of distance depends on the time delay estimation of acoustic signal, NLOS propagation often brings great error to the estimation of distance, which makes the traditional localization algorithm based on Line-of-SightLos environment invalid, resulting in the rapid decline of location accuracy. In this paper, a non-line-of-sight location method based on NLOS discrimination and a localization algorithm based on the statistical characteristics of acoustic signal channel NLOS are proposed for the indoor NLOSP-Los complex environment. The main contents and contributions of this paper are as follows: firstly, based on the indoor sound propagation model, the channel characteristics of acoustic signals in Los and NLOS environments are studied. In view of the factors such as strong multipath transmission and Doppler frequency shift in complex indoor environment, a cross-correlation-based method for estimating the relative parameters of acoustic channels (relative delay and relative gain) is proposed, which reduces the influence of Doppler effect. The computational complexity is reduced. An adaptive threshold function based on signal-to-noise ratio (SNR) is proposed to distinguish the first arrival time of acoustic signal and reduce the influence of multipath transmission. Secondly, based on the estimation of acoustic channel parameters, the characteristic extraction of channel characteristics is studied, including delay characteristic, waveform shape characteristic and RicianK coefficient, as well as amplitude distribution characteristics of gain. Nine characteristic values are used for NLOS recognition. A classifier based on support vector machine (SVM) is used to realize the recognition and classification of NLOS signals, and the combination of kernel function and eigenvalue is selected optimally. Finally, on the basis of NLOS signal recognition, the location strategy based on NLOS recognition and elimination and the weighted least square location strategy based on NLOS posteriori probability are proposed for NLOS localization of static targets. A modified Kalman filter tracking algorithm based on NLOS recognition and a modified Kalman filter tracking algorithm based on NLOS posteriori probability are proposed to achieve robust location tracking in NLOS- Los hybrid environment.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN912.3

【参考文献】