Abstract:

Imaging Photoplethysmography（iPPG） is a non-contact physiological parameter detection method using videos, but the performance of this method is easily affected by external conditions, especially sensitive to changes of ambient light and head movements. Therefore, in order to eliminate the influence of noise, this paper proposes a method based on adaptive denoising of related energy entropy threshold of Variational Modal Decomposition（VMD） and Variance Characterization Series（VCS） for obtaining anti-interference and non-contact physiological parameters according to the time-frequency characteristics of noise and iPPG signals. Firstly, Variational Modal Decomposition（VMD） method is used to decompose the iPPG signal obtained from the face video image into a series of Intrinsic Mode Functions（IMF） with certain bandwidth and frequency from high frequency to low frequency, and then the noise energy entropy of each sub-interval of the intrinsic mode functions is calculated respectively to obtain the threshold value of the component. Next, the reconstructed denoising signal is obtained by threshold processing. Finally, Variance Characterization Series（VCS） is used to detect the quality of the reconstructed signal, so as to adaptively select different signal analysis methods and realize accurate estimation of physiological parameters of heart rate and respiration rate. The proposed method is validated by the public dataset and the self-collected dataset. The experimental results demonstrate that the proposed method has stronger robustness to environmental light interference and head movements, and can obtain more accurate estimates of heart rate and respiration rate compared with the existing method.

Key words: imaging photoplethysmography, variational mode decomposition, energy entropy, adaptive de-noising, non-contact measurement, heart rate, respiratory rate

摘要：

成像式光电容积描记法（imaging Photoplethysmography，iPPG）是一种利用视频信号实现非接触式生理参数检测的方法，但该方法的性能易受外界条件影响，尤其对环境光变化及受试者头部运动等噪声敏感。为了消除噪声影响，根据噪声信号和iPPG信号的时频特性，提出了一种基于变分模态分解（Variational Mode Decomposition，VMD）的相关能量熵阈值自适应去噪与方差表征序列（Variance Characterization Series，VCS）的抗干扰非接触式生理参数信号获取方法。利用分解算法将由人脸视频中获取到的血容量脉冲信号分解为频率由高频到低频且具有一定带宽的模态分量。分别计算各个模态分量子区间的噪声能量熵从而获得该分量的阈值，再经阈值处理得到重构的去噪信号。利用VCS方法检测重构后的信号的质量从而自适应地选择信号分析方法，实现心率和呼吸率的生理参数的精确估计。在公开数据集和自采数据集上进行验证实验，实验结果表明，提出的方法对环境光与头部运动干扰有着较强的稳健性，相比目前已有的方法能够得到更精确的心率与呼吸率估计值。

关键词: 成像式光电容积描记法, 变分模态分解法, 能量熵, 自适应去噪, 非接触测量, 心率, 呼吸率