基于欧拉影像放大的非接触式心率测量方法

针对目前非接触式心率测量存在操作不便、心率同频段噪声干扰大和受环境温度影响较大等问题，提出一种基于欧拉影像放大技术的非接触式心率测量方法。首先，运用欧拉影像放大技术实现手腕处桡动脉微小跳动的动作放大；然后，对脉搏跳动视频帧的像素点亮度值在时域上进行亮度方差统计，同时在YCrCb颜色空间中分割出皮肤区域；其次，根据亮度方差统计和皮肤分割结果，结合图像形态学处理方法提取视频中桡动脉跳动区域；最后，对所提取桡动脉部位时域上亮度信号采用傅里叶变换进行时频分析，实现心率非接触式测量。实验结果表明该方法与独立成分分析法（ICA）和脉搏交流信号分析法相比，均方根误差（RMSE）分别下降50.5%和32.6%；与小波滤波法相比，平均绝对误差下降12%。非接触式心率测量方法测量结果与脉搏血氧仪测量法具有很好一致性，满足中国医药行业标准，可用于日常保健和远程医疗的心率监测。

Non-contact heart rate measurement method based on Eulerian video magnification

Aiming at problems of inconvenient operations, large noise interference of same frequency band in heart rate, and great influence by environmental temperature in existing non-contact measurement of heart rate, a non-contact measurement method of heart rate based on Eulerian video magnification technology was proposed. Firstly, a tiny beating of radial artery of wrist was magnified by Eulerian video magnification technology. Secondly, a statistics of luminance variance for pixels in an enlarged video frame was performed in time domain. Meanwhile, skin area in YCrCb color space was segmented. Thirdly, the pulsing region of radial artery in a video was extracted by luminance variance statistics and skin segmentation incorporating with image morphological processing. Finally, the non-contact measurement of heart rate was implemented by time-frequency analysis through Fourier transform of luminance signal of radial artery extracted in time domain. The experimental results show that Root Mean Square Error (RMSE) is reduced by 50.5% and 32.6%, respectively compared to Independent Component Analysis (ICA) and pulse alternating current signal analysis; Mean Absolute Difference (MAD) is 12% lower than wavelet filtering method. In this paper, the proposed approach for non-contact measurement of heart rate has a good consistency with pulse oximeter measurement, which is satisfied to pharmaceutical industry standards. It also can be used in daily family health care and telemedicine to monitor heart rate.