本文为荷兰代尔夫特大学（作者：Bruhtesfa E. Godana）的硕士论文，共100页。

基于步态特征的人体活动特征识别在医疗、养老、警察、生物特征识别等领域有着重要的应用。摄像机和人体加速度计被广泛应用于人体活动的特征识别。尽管如此，为了获得广泛的接受，用于这些应用的传感器应该不引人注目，人们的隐私应该得到尊重。

雷达是一种很有前途的人类活动特征化技术，因为它满足这些要求，并且在很大程度上不受环境条件和障碍物的影响。本文提出了室内环境下雷达信号处理和人体活动量化、识别与分类的算法。提出并讨论了从雷达数据中提取人体运动参数如活动水平、位移、速度、步长、速度带宽等的方法。利用通用软件无线电外设和GNU无线电在一个低成本的实验雷达样机上对所提出的算法进行了评估。

Human activity characterization based on gait signatures is important in several applications including medical, elderly care, police, biometric identification, etc. Cameras and on-body accelerometers are widely investigated for characterizing human activity. Nonetheless, in order to achieve wide acceptance, sensors to be used for these applications should be unobtrusive and people’s privacy should be respected. Radar is a promising technology for human activity characterization as it satisfies these requirements and operates largely unaffected by environmental conditions and the presence of blocking objects. This thesis presents algorithms for radar signal processing and human activity quantification, identification and classification in indoor environment. Methods to extract parameters of human motion such as activity level, displacement, velocity, step length, velocity bandwidth, etc., from radar data are suggested and discussed. The developed algorithms are evaluated in an experimental low-cost radar prototype using Universal Software Radio Peripheral and GNU Radio.

1.       引言

2. 人体运动特征

3. 室内环境的多普勒频移

4. 时频估计

5. 参数提取

6. 实验平台

7. 评估与讨论

8. 结论与未来工作展望

附录A USRP无线电

附录B GNU无线电

更多精彩文章请关注公众号：