Sleep/wake classification via remote PPG signals.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Remote sleep/wake classification is crucial for non-invasive health monitoring.
• Vision-based heart rate (HR) estimation often suffers from low temporal resolution.
• Existing methods may not effectively handle noise in remote photoplethysmogram (PPG) signals.

Purpose of the Study:

• To develop a remote sleep/wake classification method using high-temporal-resolution remote PPG signals.
• To introduce a dynamic HR filter for noise reduction in remote PPG signals.
• To evaluate the effectiveness of the proposed method compared to traditional approaches.

Main Methods:

• Utilizing convolutional neural networks (CNNs) to process remote PPG signals.
• Implementing a novel dynamic HR filter to mitigate noise in PPG data.
• Comparing the performance of the dynamic filter against a static filter.
• Assessing classification accuracy using the area under the ROC curve (AUC).

Main Results:

• The dynamic HR filter demonstrated superior performance over the static filter.
• The proposed remote sleep/wake classification method achieved an AUC of 0.70.
• Performance was comparable to HR estimation from wearable sensors (AUC of 0.71).

Conclusions:

• Remote PPG signals combined with CNNs offer a viable approach for sleep/wake classification.
• The dynamic HR filter significantly improves the robustness and accuracy of remote PPG-based analysis.
• This non-invasive method shows potential for sleep monitoring applications.