IoT-based Women's Safety Monitoring in Educational Campuses using WSN and Deep Learning

The recognition of human activity is significant in such domains as smart environments, security, and health. A novel approach to IoT-based women's safety monitoring at educational campuses using a hybrid deep learning model that combines CNNs and LSTMs has been proposed in this study. Using wearable inertial sensors, the system classifies six predetermined activities - walking, walking upstairs, walking downstairs, sitting, standing, and laying. The Human Activity Recognition dataset preprocessed with noise filtering and feature extraction was utilized in training and testing. Spatial features are extracted from CNNs while the LSTMs capture temporal dependencies, giving a 95.38% accuracy that is augmented with F1-score, precision, and recall exceeding 95%. This approach offers a robust, scalable safety solution for monitoring of activities that would be accomplished in real- world educational settings.