文章摘要

文章信息/Info

Title:: Non-contact heart rate detection method based on multi-region analysis using thermal infrared videos

Author(s):: LU Lei; CHENG Juan; Department of Biomedical Engineering,Hefei University of Technology,Hefei 230009, China

Keywords:: Heart rate detection; Non-contact; Multi-region analysis; Independent component analysis; Thermal infrared video

摘要:: 本研究提出了一种基于面部多区域分析的非接触式热红外视频心率检测方法。首先，确定面部3个感兴趣区域（region of interests，ROIs），构建像素均值时间序列。其次，对3个ROIs采用独立成分分析和多变量经验分解算法，分别提取包含心率信息的独立分量和本征模态函数，通过功率谱分析确定最佳独立分量和最佳本征模态函数，从而得到心率检测值。实验结果表明基于多区域分析方法的有效性，其中基于多区域独立成分分析方法的检测性能最优，平均绝对误差3.17 bpm，均方根误差2.93 bpm，标准差4.3 bpm，相关性达到了0.87，为基于视频的非接触式心率连续检测提供了解决方案。

Abstract:: We proposed a non-contact heart rate(HR) detection method based on facial multi-region analysis using thermal infrared (TIR) videos. First, three facial regions of interest (ROIs) were determined and the mean pixel values of each ROI were calculated and concatenated to form a time sequence. Then, two methods of multi-region independent component analysis (termed as MRICA) and multivariate empirical mode decomposition (termed as MEMD) were used to extract independent components (ICs) and intrinsic mode functions (IMFs). Finally, the target IC and the target IMF were determined by power spectrum analysis， and the HR was evaluated. The experimental results verified the feasibility and superiority of this method. Among them, MRICA achieved the best results with the average absolute error of 3.17 bpm, the root means square error of 2.93 bpm, the standard deviation of 4.3 bpm and the correlation of 0.87.This method provides a solution for video-based continuous non-contact HR detection applications.