This paper addresses the limitations of existing WiFi signal-based human activity recognition systems, which lack a quantitative model to correlate Channel State Information (CSI) dynamics with human activities. The authors propose CARM (CSI-based Activity Recognition and Monitoring), a system that uses two Commercial Off-The-Shelf (COTS) WiFi devices to continuously send and receive signals. CARM is based on two theoretical models: the CSI-speed model, which quantifies the correlation between CSI value dynamics and human movement speeds, and the CSI-activity model, which quantifies the correlation between movement speeds of different body parts and specific human activities. These models enable CARM to build a quantitative correlation between CSI value dynamics and specific human activities, allowing it to recognize activities by matching them to the best-fit profile. The system uses Discrete Wavelet Transform (DWT) for feature extraction and Principal Component Analysis (PCA) for denoising CSI values. Experiments show that CARM achieves an average accuracy of 96% in recognizing human activities in various environments. The paper also discusses the robustness of the models to different human activities and environmental changes, demonstrating the effectiveness of CARM in recognizing activities with varying speeds and directions.This paper addresses the limitations of existing WiFi signal-based human activity recognition systems, which lack a quantitative model to correlate Channel State Information (CSI) dynamics with human activities. The authors propose CARM (CSI-based Activity Recognition and Monitoring), a system that uses two Commercial Off-The-Shelf (COTS) WiFi devices to continuously send and receive signals. CARM is based on two theoretical models: the CSI-speed model, which quantifies the correlation between CSI value dynamics and human movement speeds, and the CSI-activity model, which quantifies the correlation between movement speeds of different body parts and specific human activities. These models enable CARM to build a quantitative correlation between CSI value dynamics and specific human activities, allowing it to recognize activities by matching them to the best-fit profile. The system uses Discrete Wavelet Transform (DWT) for feature extraction and Principal Component Analysis (PCA) for denoising CSI values. Experiments show that CARM achieves an average accuracy of 96% in recognizing human activities in various environments. The paper also discusses the robustness of the models to different human activities and environmental changes, demonstrating the effectiveness of CARM in recognizing activities with varying speeds and directions.