Lightweight Regularized Multi - Label Indoor Human Activity Recognition with CSI Fingerprints

In the rapid advancement of smart cities and smart homes, indoor human activity recognition (HAR) has gained significant importance for applications in security monitoring, health monitoring, and smart home control. This paper presents a lightweight regularized multi-label HAR method using Channel State Information (CSI) from Wi-Fi signals. We implement and evaluate a CNN-2D model enhanced with squeeze-and-Excitation (SE) module, designed for activity recognition and indoor localization in multi-user environments. The model utilizes adaptive attention mechanisms and regularization techniques to efficiently handle complex multi-label learning tasks. Comprehensive experiments on the WiMANS dataset demonstrate that our model achieves superior accuracy and efficiency, making it suitable for deployment in resource-constrained environments such as edge devices and IoT sensors. This work provides a practical solution for scalable, real-time HAR applications in smart environments.