Heart Rate Detection Based on Computer Vision

In order to measure heart rate remotely under daily scenes from face videos, this paper proposes a well-performing framework, which includes a regional strategy and KCF tracking algorithm. The former chooses regions with little noise and rich information for detection, and the latter keeps the consistency of ROIs. In addition, noises caused by light and other environmental changes are reduced by using a detrending algorithm. After that, temporal filters are used to further reduce noises caused during the process of measurement. In this paper, human face videos are collected under normal illumination, and the results are analysed quantitatively. The proposed method outperforms than other methods in accuracy and robustness.