摘要
为了实现基于IPPG(Image Photoplethysmography)技术多种环境的非接触式心率的测量,针对视频采集图像时,光照射皮肤所产生的色调、饱和度等差异造成的影响,提出 自适应HSV阈值设置的皮肤像素选择算法sA-HSV(Self-Adaptive HSV).摄像头采集整个面部区域,通过sA-HSV算法设定阈值筛选皮肤像素作为特征区域,像素点绿色通道的均值作为原始信号,加入滤波算法得到最终信号,使用傅里叶变换估计心率.仿真结果同真实仪器测量结果相比,自采集数据在LED灯光下平均绝对误差小于2.29 bpm,均方根误差小于3.67 bpm,自然光场景下平均绝对误差小于1.7 bpm,均方根误差小于2.39 bpm,公开数据集VIPL-HR下,平均绝对误差为2.57 bpm,均方根误差为3.25 bpm,该算法更适合于光照环境下的非接触式心率估计.
Abstract
In order to achieve non-contact heart rate measurement in multiple environments based on IPPG technology,aiming at the difference in hue and saturation caused by light irradiating the skin will have an impact when capturing images in a video,this paper proposes a skin pixel selection algorithm sA-HSV(Self-Adaptive HSV)with adaptive HSV threshold setting.The ordinary camera was used to collect the entire facial area as the initial feature area and set the threshold to filter the skin pixels as the characteristic area through the sA-HSV algorithm.The mean value of the green channel of the pixels was used as the original signal and added the filtering algorithm to get the final signal.The heart rate was estimated through the Fourier transform.The experimental results were compared with the standard measurement results.The average absolute error is less than 2.29 bpm and the root mean square error is less than 3.67 bpm in the self-collected datasets.The average absolute error is 2.57 bpm and the root mean square error is 3.25 bpm in the public datasets VIPL-HR.This algorithm is more suitable for non-contact heart rate measurement in illuminated environments.
维普
万方数据