<p>Reliable fall detection is essential for ensuring the safety of older adults in smart healthcare environments. Existing vision-based deep-learning fall detection systems often suffer from two major issues: high computational cost and susceptibility to false detections across varied environments. To address these problems, we propose 3D-LiteNet, a novel lightweight 3D convolutional neural network (CNN) model derived from 3D MobilenetV2, specifically optimized for fall detection in video sequences. The innovation of the proposed approach lies in its ability to reduce model parameters by over 91% compared to the baseline model, while maintaining competitive performance. Furthermore, it incorporates and systematically evaluates multiple spatiotemporal data augmentation configurations, thereby improving the model’s robustness across diverse conditions. We presented the comparison results, based on several evaluation criteria. Two standard fall detection datasets, URFD and Multicam, are used to test the proposed approach. The model achieves strong performance across multiple evaluation metrics while significantly reducing computational requirements. These findings demonstrate that 3D-LiteNet is a feasible and efficient solution for real-time fall detection in resource-limited smart healthcare systems.</p>