摘要
提出并验证了一种基于手掌毫米波雷达回波信号的身份识别技术。使用60 GHz毫米波雷达持续监测手掌回波信号,再经过目标检测、距离检测、静止检测等步骤获取符合识别要求的单次雷达回波信号,随后对这个信号进行分段滤波取平均(PNA)的数据预处理,最后送入由一维卷积层为核心的神经网络(RNet)分类器中进行识别,得到被检测者的身份信息。本研究采集了包含120000个样本的数据集,并在此之上验证了数据预处理算法和分类识别算法,最终取得96.8%的综合识别准确率。由于采用雷达回波信号作为身份信息的表征,相比于传统身份识别方法无光照条件要求、无需直接接触传感器且无隐私泄露风险,拥有广阔的应用场景。
This thesis presents and verifies an identity recognition technology based on hand millimeter wave radar echo signal.The technique involves placing the user′s hand flat in front of the radar.Radar echo signal,through the analysis of the real-time target detection,distance test,static test steps such as access to meet the requirements of the identification of single radar echo signal,then the signal block filter by averaging(PNA)data preprocessing,finally will handle good data into the one-dimensional convolution as the core layer of neural network reasoning(RNet)classifier,get the user′s identity information.In this thesis,a data set containing 120000 samples is established,on which the data preprocessing algorithm and classification recognition algorithm are verified,and the comprehensive recognition accuracy of 96.8%is finally achieved.Because radar echo signal is used as the representation of identity information,compared with the traditional identification method,it has a wide range of application scenarios with no illumination requirements,no direct contact with the sensor and no risk of privacy leakage.
作者
曹佳禾
陈君毅
王智铭
蒋德琛
王勇
Cao Jiahe;Chen Junyi;Wang Zhiming;Jiang Dechen;Wang Yong(College of Information Science&Electronic Engineering,Zhejiang University,Hangzhou 310013,China)
出处
《国外电子测量技术》
北大核心
2022年第3期170-176,共7页
Foreign Electronic Measurement Technology
关键词
毫米波雷达
身份识别
深度学习
millimeter-wave radar
identification
deep learning
