An IoT-enabled CRNN framework for secure wearable sensor-based activity recognition in physical education.

Area of Science:

• Computer Science
• Biomedical Engineering
• Internet of Things

Background:

• Digital technologies and wearable sensors enable large-scale physiological and kinematic data acquisition.
• Accurate human activity recognition is challenging due to spatiotemporal dependencies in sensor data.
• Existing methods struggle to effectively model complex patterns from multimodal wearable sensor streams.

Purpose of the Study:

• To propose a novel IoT-oriented activity recognition framework using a Convolutional Recurrent Neural Network (CRNN).
• To effectively model spatial and temporal characteristics of multimodal wearable-sensor data streams for enhanced activity classification.
• To demonstrate the framework's superiority over conventional and deep learning methods.

Main Methods:

• Utilized multidimensional wearable sensor datasets including physiological and inertial measurements.
• Applied preprocessing and temporal segmentation to raw sensor signals.
• Employed a CRNN architecture integrating Convolutional Neural Networks (CNNs) for spatial feature extraction and Recurrent Neural Networks (RNNs) for temporal modeling.

Main Results:

• The proposed CRNN framework achieved 98.2% classification accuracy, 97.2% sensitivity, 99.2% specificity, 97.4% recall, and 97.6% precision.
• Demonstrated superior performance compared to five representative baseline methods.
• Showcased higher recognition accuracy and greater generalization robustness.

Conclusions:

• The CRNN framework offers an effective solution for wearable-sensor-based activity recognition.
• The model shows significant potential for deployment in IoT-enabled monitoring and educational settings.
• The synergistic integration of CNNs and RNNs effectively addresses spatiotemporal dependencies in sensor data.