Towards Wearable Respiration Monitoring: 1D-CRNN-Based Breathing Detection in Smart Textiles.

Area of Science:

• Wearable technology
• Biomedical engineering
• Machine learning for healthcare

Background:

• Respiratory activity monitoring is crucial for physiological health assessment.
• Smart e-textiles offer unobtrusive vital sign monitoring capabilities.
• Accurate classification of breathing patterns is essential for wearable health systems.

Purpose of the Study:

• To develop and evaluate a 1D-CRNN for automatic breathing activity classification.
• To assess the performance of the 1D-CRNN using inertial data from smart e-textiles.
• To demonstrate the near-real-time feasibility of the proposed method for wearable applications.

Main Methods:

• Utilized a 1D-CRNN integrating convolutional and recurrent layers for feature extraction and temporal modeling.
• Trained and evaluated the model on inertial data from 59 subjects using stratified five-fold cross-validation.
• Assessed performance across different window sizes, focusing on accuracy and F1-score.

Main Results:

• Achieved a mean accuracy of 0.88 and an F1-score of 0.92 at a 2000-sample window size.
• The best single-fold configuration reached an accuracy of 0.995 and an F1-score of 0.99.
• Demonstrated near-real-time processing (1.76s for 250s measurement), over 100x faster than recording time.

Conclusions:

• The 1D-CRNN effectively classifies breathing activity from e-textile inertial data.
• The proposed approach is suitable for embedded, on-device inference in wearable systems.
• High accuracy and real-time performance enable practical applications in continuous health monitoring.