Alam Shafaf, Singh Surya P N, Abeyratne Udantha
Annu Int Conf IEEE Eng Med Biol Soc. 2017 Jul;2017:4293-4296. doi: 10.1109/EMBC.2017.8037805.
Respiratory rate can be a vital indicator of illness; however, tracking this is a non-trivial process. Phase-based Eulerian Video Magnification (EVM) is an exciting spatiotemporal video processing approach able to reveal subtle breathing motions within video sequences; however, its results are variant to large motions and camera blur. In the case of camera motion, a compensation strategy of stabilizing (without smoothing) the video has the may reduce estimation error in handheld cases. This work explores the extent of removing motion artefacts and its impact on identifying subtle breathing motions. Tests across six indoor scenes show a reduction mean breathing estimate error for 4 of 6 cases and highlights the sensitivity of this approach to unwanted body movements. The results of this project suggest the plausibility that non-smoothing video amplification processes can be an effective method to track breathing motion and that implementing correction techniques which may allow a smartphone to provide a compact, non-invasive, online breathing monitor.
呼吸频率可能是疾病的一个重要指标；然而，追踪它是一个并非易事的过程。基于相位的欧拉视频放大（EVM）是一种令人兴奋的时空视频处理方法，能够揭示视频序列中的细微呼吸运动；然而，其结果会因大幅度运动和相机模糊而变化。在相机运动的情况下，一种稳定（不进行平滑处理）视频的补偿策略可能会减少手持情况下的估计误差。这项工作探讨了去除运动伪影的程度及其对识别细微呼吸运动的影响。在六个室内场景中进行的测试表明，六个案例中有四个案例的平均呼吸估计误差有所降低，并突出了这种方法对不必要身体运动的敏感性。该项目的结果表明，非平滑视频放大过程有可能成为追踪呼吸运动的有效方法，并且实施校正技术可能使智能手机能够提供紧凑、无创的在线呼吸监测器。
