Integrated Respiratory Rate Detection for Vehicle Drivers using Multi-Modal Signal Fusion

Accurate monitoring of the respiratory rate (RR) is crucial for early disease detection and effective health assessment. Real-world applications often face challenges such as motion artifacts and environmental noise. Robust signal fusion techniques are essential to enhance data reliability and to ensure that signals are accurate despite external disturbances.To address these challenges, we present an integrated respiratory rate detection system for vehicle drivers, leveraging convolutional neural networks (CNN) and bidirectional long short-term memory (BiLSTM)-based models for non-invasive health monitoring. We detect the RR with the four sensors: two accelerometers, a piezoelectric sensor, and an RGB camera. We use a chest belt to record the ground truth. The system tracks the driver's RR and handles challenges such as motion artifacts and environmental noise. Tested on 15 participants, the BiLSTM-based signal fusion model achieves F₁ = 0.68, demonstrating its effectiveness in detecting respiratory patterns. The mobile interface visualizes health data and uses color-coded alerts.In contrast to wearables, in-vehicle health recording supports individualized long-term monitoring without any user interaction and enhances data stability. Potential applications are preventive healthcare, remote monitoring, and telemedicine. Future work will focus on optimizing sensor fusion and expanding the model's capabilities to support health analytics in intelligent transportation systems.