Scholarlink

<p>Blood oxygen saturation (SpO<sub>2</sub>) is a critical indicator of lung function, and its convenient and rapid monitoring is vital for preventing various diseases. Recent advancements in non-contact SpO<sub>2</sub> measurement using RGB cameras have demonstrated its potential in diverse application scenarios. This paper presents a method to extract three-channel pulse signals from designated facial video regions and estimates blood oxygen saturation signals using traditional regression machine learning models. By analyzing facial videos, this paper approach aims to achieve accurate and non-contact SpO<sub>2</sub> monitoring, enhancing accessibility in remote and virtual healthcare environments. We conducted experiments on publicly available datasets, demonstrating that paper method can precisely estimate SpO<sub>2</sub> from facial videos of different subjects. The proposed features exhibit excellent performance across multiple machine learning regression models, indicating good research value and applicability in fields such as health screening and telemedicine. The relevant code can be accessed at: <a href="https://github.com/Si-tong-416/my_first_code_to_pre_SpO2">https://github.com/Si-tong-416/my_first_code_to_pre_SpO2</a></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Non-contact SpO2 monitoring via multi-channel pulse signals from facial videos using machine learning

  • 期刊论文

  • 发表日期：

    2025年9月29日

查看全文
  • Tong Si,
  • Yuanbo Song,
  • Peng Sun,
  • Jinyuan Xiao,
  • Rui Zou,
  • Zhencheng Chen

摘要

Blood oxygen saturation (SpO2) is a critical indicator of lung function, and its convenient and rapid monitoring is vital for preventing various diseases. Recent advancements in non-contact SpO2 measurement using RGB cameras have demonstrated its potential in diverse application scenarios. This paper presents a method to extract three-channel pulse signals from designated facial video regions and estimates blood oxygen saturation signals using traditional regression machine learning models. By analyzing facial videos, this paper approach aims to achieve accurate and non-contact SpO2 monitoring, enhancing accessibility in remote and virtual healthcare environments. We conducted experiments on publicly available datasets, demonstrating that paper method can precisely estimate SpO2 from facial videos of different subjects. The proposed features exhibit excellent performance across multiple machine learning regression models, indicating good research value and applicability in fields such as health screening and telemedicine. The relevant code can be accessed at: https://github.com/Si-tong-416/my_first_code_to_pre_SpO2