基于MFCC特征的Wi-Fi信道状态信息人体行为识别方法  

HUMAM BEHAVIOR RECOGNITION METHOD BY WI-FI CHANNEL STATE INFORMATION BASED ON MFCC CHARACTERISTICS

作　　者：蒙倩霞 余江[1] 常俊[1] 浦钰 Meng Qianxia;Yu Jiang;Chang Jun;Pu Yu(School of Information Science and Engineering,Yunnan University,Kunming 650500,Yunnan,China)

机构地区：[1]云南大学信息学院,云南昆明650500

出　　处：《计算机应用与软件》2022年第12期125-131,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(61162004);云南省教育厅科学研究基金项目(2019J0007);云南大学研究生科研创新基金项目(2019151)。

摘　　要：CSI(Channel State Information)可提供被动的人体行为识别方法,根据CSI和声音信号传播相似性和共享频谱带宽的特性将MFCC(Mel-Frequency Cepstral Coefficients)特征提取应用于CSI时间序列,并对复杂场景下的视距和非视距的几种日常行为进行识别。该方法对数据去噪、PCA、相位校准处理,从预处理后的信号中提取了MFCC统计特征和一个无偏移对数频谱能量,用蚁群和粒子群混合优化SVM进行分类识别。实验结果表明,该方法能有效识别复杂场景下的日常行为,在视距情况下,平均识别率达到了91%。Channel state information(CSI) can provide passive human behavior recognition methods. According to the characteristics of CSI and sound signal propagation similarity and shared spectrum bandwidth, Mel-frequency cepstral coefficients(MFCC) feature extraction was applied to CSI time series, and several daily behaviors of LOS and NLOS in complex scenes were identified. The data denoising, PCA and phase calibration were processed. The MFCC statistical features and an unbiased logarithmic spectrum energy were extracted from the preprocessed signal, and the SVM based on ant colony and particle swarm optimization was used for classification and recognition. The experimental results show that this method can effectively recognize the daily behavior in complex scenes, and the average recognition rate is 91% in the case of LOS distance.

关 键 词：CSI 行为识别 PCA MFCC 支持向量机 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术] TN99[电子电信—信号与信息处理]