穿墙雷达人体行为识别理论与技术

This project mainly deals with time-varying behavior recognition problem of the human behind the walls for through-the-wall radar. The theory and technology can be established and improved by investigating the following fundamental problems: human typical behaviors and echo signal modeling for through-the-wall radar, human behavior recognition architecture based on multi-channel feature spectrum, real-time identification of time-varying human behavior. This project will propose multi-frequency distribution diversity based multipath suppression methods to remove these complex ghosts, and the recurrent neural network (RNN) based recognition approaches to identify different human behaviors, which can solve the feature extraction of the human echo in the enclosed building with limited data as well as real-time recognition of the complex human behaviors with unknown starting points and the time interval. This project will build an experimental system and carry out a series of verification tests for these proposed theories and technical achievements, which can provide theoretical and technical support for the development of system equipment. The research is of great scientific significance, since it can improve the fundamental research levels and the innovational abilities of our country for the through-the-wall radar based human behavior recognition. Moreover, the theory results and techniques can be applied to several fields, such as antiterrorism, street battle, disaster rescue, and patient care.

本项目旨在解决穿墙雷达人体目标的时变行为识别问题，开展穿墙雷达人体典型行为与回波信号建模、基于多通道特征谱图的识别架构、时变行为实时识别等技术研究，提出基于多频点分布差异的多径抑制以及基于循环神经网络（RNN）的行为识别等方法，解决了封闭空间有限样本人体行为回波特征提取和非定时长和非定起始时刻的复杂行为实时识别等两大关键科学问题，建立和完善穿墙雷达人体行为识别理论与技术体系；搭建试验系统，对提出的相关理论和技术成果进行实验验证，为系统装备研制提供理论与技术支撑。本项目研究可提高我国在穿墙雷达人体行为识别领域的基础研究水平和自主创新能力，具有重要的科学意义；其理论成果和技术可应用于反恐维稳、城市巷战、灾害救援和病人监护等领域。

本项目旨在解决穿墙雷达人体行为识别问题，通过研究穿墙雷达典型人体行为与回波建模、人体行为回波杂波抑制及特征构建、穿墙雷达人体行为识别方法等内容，提出了优化距离像、离散时变特征空间点、多频融合的CEEMADAN时频谱图等人体行为表征方法，以及基于轻量级网络、多域特征融合、多视角信息融合等人体行为识别方法，解决了封闭空间有限样本人体行为回波微多普勒关键特征提取、起始时刻未知与非定时长人体行为实时识等关键科学问题，建立和完善了穿墙雷达人体行为识别理论与技术体系；搭建了试验系统，对所提人体行为识别方法的可行性、有效性进行了实验验证。本项目研究可提高我国穿墙雷达人体行为识别领域的基础研究水平和自主创新能力，具有重要的科学意义；其理论成果和技术还可应用于医疗康复、反恐维稳、城市巷战、智慧汽车等领域。