摘要: |
为了在人体微多普勒特征不明显条件下识别静态人体目标及人体姿态,提出了一种结合双谱对角线起伏特性与目标强散射点分布特征进行人体目标识别的方法。首先,通过分析静态人体目标双谱,提取双谱对角线起伏特性作为分类特征,降低了双谱数据的维数,减少了双谱特征冗余。然后,结合目标强散射点分布特征从不同角度描述目标,并构造用于目标识别的特征向量。最后,用支持向量机实现目标识别。仿真和实测结果均表明,双谱对角线起伏特性与目标强散射点分布特征融合的方法可以有效识别出静态人体目标并且实现人体姿态识别。 |
关键词: 超宽带雷达 人体目标识别 双谱起伏特性 支持向量机 |
DOI: |
|
基金项目:国家自然科学基金资助项目(61162007);广西自然科学基金资助项目(2013GXNSFAA019323);广西科学研究与技术开发计划项目(桂科攻14122006-6);广西教育厅科研立项项目(KY2015LX096) |
|
Ultra-wideband radar human target recognition based on bispectrum feature |
JIANG Liubing,JI Yawen,YANG Tao,CHE Li |
() |
Abstract: |
In order to identify human targets when micro-Doppler features are not obvious,a method combining diagonal fluctuation characteristic of echo bispectrum with distribution of strong scattering points is presented. Firstly,the bispectrum of static human targets is analyzed and diagonal fluctuation characteristic of echo bispectrum is extracted. Thus,characteristic dimension of bispectrum and feature redundancy are reduced. Secondly,to describe targets in different aspects,distribution feature of strong scattering points is combined with bispectrum and feature vector which is used to recognize targets is constructed. Finally,targets recognition is realized by using Support Vector Machine(SVM). Both simulation and measurement results show that the proposed method can effectively recognize human target and human posture. |
Key words: ultra-wideband radar human target recognition bispectrum fluctuation characteristic support vector machine |