非合作面部晃动情况下的心率检测

发布时间：2018-09-04 18:27

【摘要】：目的心率是直接反映人体健康的重要指标之一,基于视频的非接触式心率检测在医疗健康领域具有广泛的应用前景。然而,现有的基于视频的方法不适用于复杂的现实场景,主要原因是没有考虑视频中目标晃动干扰和空间尺度特征,使得血液容积脉冲信号提取不准确,检测精度不尽人意。为了克服以上缺陷,提出一种抗人脸晃动干扰的非接触式心率检测方法。方法本文方法主要包含3个步骤:首先,针对目标晃动干扰人脸区域选择的问题,利用判别响应图拟合检测参考图像的人脸区域及主要器官特征点,在人脸跟踪时首次引入倾斜校正思想,输出晃动干扰抑制后的人脸视频;然后,结合空间尺度的差异,采用颜色放大方法对晃动干扰抑制后的人脸视频进行时空处理,提取干净的血液容积脉冲信号;最后,考虑到小样本问题,通过傅里叶系数迭代插值的频域分析方法估计心率。结果在人脸静止的合作情况以及人脸晃动的非合作情况下采集视频,对心率检测结果进行定量分析,本文方法在两种情况下的准确率分别为97.84%和97.30%,与经典和最新的方法相比,合作情况准确率提升大于1%,非合作情况准确率提升大于7%,表现了出色的性能。结论提出了一种基于人脸视频处理的心率检测方法,通过有效分析人脸的晃动干扰和尺度特性,提取到干净的血液容积脉冲信号,提高了心率检测的精度和鲁棒性。
[Abstract]:Objective Heart rate is one of the important indexes to reflect human health directly. Non-contact heart rate detection based on video has a wide application prospect in the field of medical health. However, the existing video-based methods are not suitable for complex real scenes, the main reason is that the target sloshing interference and spatial scale characteristics are not considered in the video, which makes the extraction of blood volume pulse signal inaccurate and the detection accuracy unsatisfactory. In order to overcome the above defects, a non-contact heart rate detection method against face sloshing interference is proposed. Methods the method consists of three steps: firstly, aiming at the problem of target sloshing interference, the discriminant response graph is used to fit the face region and the main organ feature points of the reference image. In face tracking, tilt correction is introduced for the first time to output face video after swinging interference suppression. Then, color amplification method is used to process face video after sloshing interference suppression by combining spatial scale difference. The clean blood volume pulse signal is extracted. Finally, considering the small sample problem, the heart rate is estimated by Fourier coefficient iterative interpolation method in frequency domain. Results the video was collected under the condition of static face cooperation and the non-cooperative situation of face sloshing, and the results of heart rate detection were quantitatively analyzed. The accuracy of the two methods were 97.84% and 97.30% respectively, compared with the classical method and the latest method. The accuracy of cooperation is greater than 1, and that of non-cooperation is more than 7, showing excellent performance. Conclusion A heart rate detection method based on face video processing is proposed. By analyzing the sloshing interference and scale characteristics of face effectively, the clean blood volume pulse signal is extracted, which improves the accuracy and robustness of heart rate detection.
【作者单位】：合肥工业大学计算机与信息学院;
【基金】：国家自然科学基金项目(61175033,61503111) 安徽省自然科学基金项目(1508085SMF222) 合肥工业大学应用科技成果培育计划(JZ2016YYPY0051)~~
【分类号】：R540.4;TP391.41

【相似文献】