Machine learning models for camera-based physiological measurement can have weak generalization due to a lack of representative training data. Body motion is one of the most significant sources of noise when attempting to recover the subtle cardiac pulse from a video. We explore motion transfer as a form of data augmentation to introduce motion variation while preserving physiological changes. We adapt a neural video synthesis approach to augment videos for the task of remote photoplethysmography (PPG) and study the effects of motion augmentation with respect to 1) the magnitude and 2) the type of motion. After training on motion-augmented versions of publicly available datasets, the presented inter-dataset results on five benchmark datasets show improvements of up to 75% over existing state-of-the-art results. Our findings illustrate the utility of motion transfer as a data augmentation technique for improving the generalization of models for camera-based physiological sensing. We release our code and pre-trained models for using motion transfer as a data augmentation technique on our project page: https://motion-matters.github.io/

翻译：针对摄像机式生理测量的机器学习模型常因缺乏代表性训练数据而导致泛化能力不足。当试图从视频中恢复细微的心跳脉冲时，身体运动是最显著的噪声源之一。本文探索将运动迁移作为一种数据增强手段，在保留生理变化的同时引入运动多样性。我们采用神经视频合成方法对视频进行增强，以服务于远程光电容积描记术（PPG）任务，并研究运动增强在以下两个维度的效果：1)运动幅度，2)运动类型。在公开数据集的运动增强版本上进行训练后，基于五个基准数据集的跨数据集结果表明，相较于现有最优方法，性能提升最高可达75%。我们的发现阐明了运动迁移作为数据增强技术在提升摄像机式生理感知模型泛化能力方面的价值。相关代码及预训练模型已发布在项目页面：https://motion-matters.github.io/