基于iForest与KDE的雷达目标最优距离门估计

Estimation of optimal range bin for radar target based on iForest and KDE

针对脉冲相干式毫米波雷达在人体生命体征信号检测中胸廓目标的距离门突变和切变问题,提出一种基于孤立森林(iForest)与核密度估计(KDE)的雷达目标最优距离门估计方法。为有效评估目标信号的距离门检测准确度,设计了人体呼吸模拟装置,建立目标信号跟踪与测量评估方法;基于iForest与KDE算法对目标最优距离门选取进行实验研究。结果表明,静态下误差率最高为5.17%;动态下误差率为8.30%。为进一步验证其有效性,对人体进行呼吸检测,并且提取的人体呼吸信号杂波少,信噪比高,能够有效提取呼吸信号特征。因此,基于iForest与KDE算法在目标最优距离门选取具有较好的估计能力,为该类型毫米波雷达非接触生命体征提取提供了有效的目标跟踪方法。

Aiming at the range bin mutation and shear of thoracic target in the detection of human vital signs by pulse coherent millimeter wave radar,a method for estimating the optimal range bin of radar targets based on isolation forest(iForest)and Kernel Density Estimation(KDE)was proposed.In order to effectively evaluate the detection accuracy of the range bin of the target signal,a human breathing simulation device was designed,and a target signal tracking and measurement evaluation method was established;based on the iForest and KDE algorithm,an experimental studied on the selection of the optimal range bin of the target,the results showed that the error rate was up to 5.17%under static conditions and 8.30%under dynamic conditions.In order to further verify its effectiveness,the human respiratory measurement was performed.In addition,the extracted human respiratory signal had less clutter and high signal-to-noise ratio,which can effectively extract the characteristics of the respiratory signal.Therefore,based on the iForest and KDE algorithm,it had a good estimation ability in the selection of the optimal range bin of the target,and provided an effective target tracking method for the non-contact vital signs extraction of this type of millimeter wave radar.