基于视频的非接触式心率检测的研究

发布时间：2018-01-30 09:42

本文关键词： 趋势滤波奇异谱分析独立成分分析时域差分操作　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着现代生活水平的提高,人们对于疾病预防和日常健康监测的需求也日益增长。其中,心率作为一个重要的生命体征信息,是检测心血管疾病以及指导科学锻炼的重要参数信息。所以如何方便快捷的检测出人体的心率信息,是信号处理领域新兴的研究热点。传统的心率检测仪器比如心电图等需要接触人的身体,而且价格昂贵,不适合用于日常的健康监护,而现在的非接触式心率检测方法比如多普勒雷达,热成像技术等也都需要专用的设备,成本门槛较高.所以,基于视频的非接触式心率检测的方法一经出现,就引起了广泛的关注。基于视频的非接触式心率检测方法根据光电容积脉搏波描记法的原理,只需要一个可以拍摄视频的设备,不需要直接接触人的皮肤,而是通过人脸的检测与跟踪,去趋势,独立成分分析,心率提取等算法得到心率。现有的基于视频的非接触式心率检测方法需要被测者保持静止且检测环境的光线需要保持恒定,在光线变化或者运动状态下检测结果误差很大。本文通过改善信号去趋势的方法,用L1趋势滤波代替L2趋势滤波,之后使用奇异谱分析算法来排除非心率信号的干扰,可以有效地消除由于光线变化带来的噪声。另外,通过检测被检测者的面部运动信息,并在奇异谱分析中进行相关处理,可以进一步排除视频中面部运动带来的干扰。最后我们进一步利用时域差分操作使心率信号增强的同时抑制噪声信号的干扰,可以更有效地排除运动对心率信号测量造成的干扰。实验结果表明,在静止,运动和光线变化三种情况下,本文提出的算法都有良好的检测效果。利用本文提出的算法得到的平均心率误差值在1.368BPM,可以达到实际应用的需求。
[Abstract]:With the improvement of modern living standards, the demand for disease prevention and daily health monitoring is also increasing. Among them, heart rate is an important vital sign information. It is an important parameter information to detect cardiovascular disease and guide scientific exercise. So how to detect the human heart rate information quickly and conveniently. It is a new research hotspot in signal processing field. Traditional heart rate detection instruments such as electrocardiogram need to touch human body, and the price is expensive, so it is not suitable for daily health monitoring. The current non-contact heart rate detection methods such as Doppler radar, thermal imaging technology also need special equipment, high cost threshold. Therefore, video-based non-contact heart rate detection methods have emerged. The non-contact heart rate detection method based on video, according to the principle of photovolumic pulse wave recording, only needs a device that can shoot video, and does not need direct contact with human skin. But through face detection and tracking, detrend, independent component analysis. The existing contactless heart rate detection methods based on video need the subjects to remain static and the light in the detection environment to remain constant. In this paper, L1 trend filter is used to replace L2 trend filter by improving the signal de-trend method. After that, the singular spectrum analysis algorithm is used to eliminate the interference of non-heart rate signal, which can effectively eliminate the noise caused by the change of light. In addition, the facial motion information of the detected person is detected. And in the singular spectrum analysis of correlation processing, can further eliminate the facial movement in the video interference. Finally, we further use the time-domain differential operation to enhance the heart rate signal and suppress the interference of noise signal. The interference caused by exercise on heart rate signal measurement can be eliminated more effectively. The experimental results show that there are three cases: static motion and light change. The average heart rate error obtained by the proposed algorithm is 1.368 BPM, which can meet the requirement of practical application.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R540.4;TP391.41

【相似文献】