Human activity recognition plays a significant role in smart building applications, healthcare services, and security monitoring. In particular, WiFi‐based indoor wireless sensing system becomes increasingly popular due to its noninvasiveness. This work presents the design and implementation of DARMS, a Device‐free human Activity Recognition and Monitoring System that can be deployed with low‐cost commodity WiFi devices. DARMS is a passive wireless sensing system, and it can accurately distinguish various daily activities without the user wearing any sensor. DARMS makes two key technical contributions. First, an effective signal processing methodology is designed to extract the CSI features both in the time domain and frequency domain. Second, a dual‐channel neural network that combines temporal and frequency information is proposed to achieve fine‐grained activity recognition. In our experiments, DARMS shows outstanding performance in different indoor environments, with an average accuracy of 96.9% for fall detection and 93.3% for human activity recognition.