基于网络可视图的室内人体状态检测研究

Research on indoor human body state detection based on visibility graph

随着无线局域网的普及,利用无线信号对室内人员活动状态的研究越来越多,如进行无源被动的人体朝向感知等。本文提出了一种结合无线信道状态信息(CSI)和可视图(VG)复杂网络技术的室内人体朝向检测方法,首先以无线局域网中的信道状态幅度和相位信息构建时间序列数据网络,然后基于提取的网络参数和原始统计属性作为融合特征,再通过机器学习算法进行人体朝向的分类检测。为验证算法效果,本文建立了信道状态幅度和位相信息采集平台,综合多对天线数据进行了教室和办公室2种环境下的人体4朝向和8朝向检测,还讨论了K近邻(KNN)、朴素贝叶斯(NB)和支持向量机(SVM)等分类方法的时间复杂度对比。实验结果表明,本文所提出的方法和实验方案具有较高的室内人体朝向检测精度,8朝向的最佳检测精度能达到98.66%。

Nowadays many researches have been focusing on the monitoring of indoor people activity status using wireless signals,such as passive orientation detection.This paper proposes a human body orientation detection method which combines wireless channel state information(CSI)and visual graph(VG)complex network technique.Based on the time series data network method,the channel state amplitude and information in the wireless local area network are used as the original physical quantity to construct the complex network and further extract the network features,then the machine learning algorithm is used to perform classification of the human body orientation.In order to verify the effect of the algorithm,firstly the channel state amplitude and phase information collection platform are established,then the integrated pairs of antenna data are used to detect the 4-direction and 8-direction of the human body in both classroom and office environments.Meanwhile,classification methods such as K-nearest neighbor(KNN),naive Bayes(NB)and support vector machine(SVM)are also discussed and compared on the time complexity aspects.The experimental results show that the proposed method and experimental scheme have high indoor body orientation detection accuracy,and the best detection of 8-direction can reach 98.66%.