<p>This paper introduces a novel approach for indoor activity recognition, addressing the limitations of existing methods, such as privacy concerns, high cost, the need for specialized hardware, and intrusive monitoring. Vision-based and wearable sensor-based methods often struggle with privacy issues, user discomfort, and the need for additional devices, while RF-based methods may require significant computational resources and large labeled datasets. By leveraging the ubiquity of WiFi infrastructure, we utilize WiFi Channel State Information (CSI) to capture fine-grained information about the indoor environment and human activities. Our system employs advanced signal processing and deep neural networks for accurate classification of activities. Through extensive evaluation, we demonstrate the system's robustness across various scenarios and offer insights into its practical implementation. By prioritizing user privacy and utilizing existing infrastructure, our solution offers a promising avenue for creating smart environments that can detect and recognize human activities within indoor spaces without intrusive monitoring or additional costs.</p>