Automated Sleep Analysis Pipeline Using Consumer-Accessible Hardware
Brought sleep study capabilities to consumers and researchers outside clinical environments, enabling scalable and low-cost analysis of sleep quality.
Situation
Polysomnography (PSG) sleep studies are traditionally conducted in clinical settings with expensive equipment and manual analysis, limiting accessibility and scalability.
Solution
Developed an end-to-end automated sleep analysis pipeline spanning data acquisition, signal processing, feature extraction, and analysis automation.
OUTCOMES
Challenges
Accessibility
- •Clinical-only sleep studies
- •High equipment costs
Automation
- •Manual scoring workflows
- •Limited consumer tooling
Solutions
PSG-Compatible Capture
Multi-channel biosignal capture aligned with PSG methodologies (EEG, EOG, EMG)
- Captured EEG, EOG, and EMG signals
- Aligned acquisition with PSG methodologies
- Supported consumer-accessible hardware workflows
Signal Filtering Pipeline
Notch and bandpass filtering for relevant biosignal bands.
- Removed powerline interference
- Applied notch filtering
- Isolated relevant frequency bands
- Improved signal quality for downstream analysis
Standardized Data Formats
Conversion to standardized formats (e.g., EDF)
- Converted recordings into standardized formats
- Supported EDF-compatible workflows
- Improved interoperability with scientific tooling
Spectral Analysis Engine
FFT-based spectral analysis to compute power spectral density.
- Computed power spectral density
- Segmented recordings into 30-second intervals
- Classified Delta, Theta, Alpha, and Sigma bands
- Prepared features for automated staging
Sleep Stage Automation
Automated sleep stage classification based on spectral features.
- Automated sleep stage classification
- Reduced manual analysis requirements
- Supported plugin analysis using external scientific libraries