Yang Xiuzhu, Yu Yibo, Qian Hongyu, Zhang Xinyue, Zhang Lin
Annu Int Conf IEEE Eng Med Biol Soc. 2020 Jul;2020:485-488. doi: 10.1109/EMBC44109.2020.9176528.
Utilizing Impulse Radio Ultra-WideBand (IR-UWB) radar for vital sign monitoring has attracted growing interest due to the noncontact measurement without privacy concerns. Most of existing researches assume that the subject's chest is directed to the radar antenna, to ensure the strength of backscattered signals from chest movement. However, a large angle between the antenna and the subject's chest caused by the body orientation badly affects the monitoring accuracy. Multiple observations of the same cardiopulmonary activity from different orientations provide more available measurements. This paper addresses the challenge by using an IR-UWB radar network instead of a single radar. Three IR-UWB radars are placed as endpoints of an equilateral triangle to collect vital sign information of a subject sitting at the center. A Conditional Generative Adversarial Network (CGAN) method is proposed to fuse multisensory data. First, the body orientation is classified by combining signal features and a random forest classifier. Then the impact of different angles on vital sign monitoring results is discussed and validated in each orientation. The data fusion process is modelled as an extended generative network with orientation based condition to produce the enhanced vital signal. This signal is optimized with the discriminator network to a fitted sinusoidal wave with heartbeat and respiratory information. Experimental results on measuring Heartbeat Rate (HR) in different orientations reveal the effectiveness and stability of the proposed method.
利用脉冲无线电超宽带（IR-UWB）雷达进行生命体征监测，由于其非接触式测量且不存在隐私问题，已引起越来越多的关注。现有的大多数研究都假定受试者的胸部朝向雷达天线，以确保来自胸部运动的反向散射信号的强度。然而，身体姿势导致天线与受试者胸部之间的大角度会严重影响监测精度。从不同方向对同一心肺活动进行多次观测可提供更多可用测量值。本文通过使用IR-UWB雷达网络而非单个雷达来应对这一挑战。将三个IR-UWB雷达放置在等边三角形的端点处，以收集坐在中心位置的受试者的生命体征信息。提出了一种条件生成对抗网络（CGAN）方法来融合多感官数据。首先，通过结合信号特征和随机森林分类器对身体姿势进行分类。然后在每个方向上讨论并验证不同角度对生命体征监测结果的影响。数据融合过程被建模为一个基于方向条件的扩展生成网络，以产生增强的生命信号。该信号通过判别器网络优化为带有心跳和呼吸信息的拟合正弦波。在不同方向上测量心率（HR）的实验结果揭示了所提方法的有效性和稳定性。
