High-Reliability Signal Quality Validation for Biosignals Using Sensor Fusion and Software Indices.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Biosignal quality is crucial for accurate physiological monitoring and diagnostics.
• Existing methods often struggle with artifacts from motion or poor sensor contact.
• The need for robust, automated quality assessment is growing for clinical applications and AI development.

Purpose of the Study:

• To propose and validate a two-stage hybrid framework for biosignal quality assessment.
• To enable beat-level or segment-level labeling for real-time filtering and offline dataset curation.
• To design a modular architecture extensible to various biosignals beyond ECG.

Main Methods:

• A two-stage framework combining sensor-integrity gating and software-based signal quality indices.
• Stage 1: Rejects corrupted intervals using motion sensing (IMU) and electrode status.
• Stage 2: Applies physiological plausibility, morphological consistency (DTW), SNR, and baseline wander analysis.

Main Results:

• Achieved 98.1% classification accuracy on expert-annotated ECG data.
• Obtained ~98% F1-score, 99% sensitivity, and 97% specificity.
• Demonstrated the framework's potential for improving data reliability in downstream analyses.

Conclusions:

• The proposed hybrid framework significantly enhances biosignal quality validation accuracy.
• Its modular design supports extensibility to diverse biomedical time-series signals.
• This approach promises more stable monitoring, reduced false alarms, and higher-quality datasets for AI.