Unveiling Human Behavior: Deep Learning Approaches for Activity Recognition
D. Lakshmi, D. Sattibabu, B. Sindhu · 2025
Human Activity Recognition (HAR) is an essential activity across several fields, including human-computer interaction, surveillance, and healthcare. This research employs deep learning methodologies, namely Convolutional Neural Networks (CNN) and transfer learning, to automate Human Activity Recognition (HAR) across many areas. The CNN model derives spatial and temporal properties from raw sensor data, allowing it to identify intricate activity patterns. The suggested approach attains a 95% accuracy rate in identifying several human activities, including standing, walking, jogging, and sitting. CNN, in contrast to conventional machine learning models, can process unrefined input signals, making them proficient in human activity recognition using continuous time-series data, trained with labeled datasets. Transfer learning enhances the efficiency and accuracy of a pre-trained CNN model by minimizing taskspecific data acquisition and augmenting its resilience to human activities. This system integrates CNN with transfer learning, attaining 95% accuracy, and rendering it appropriate for wearable device monitoring, smart homes, and physical rehabilitation systems.