Enhancing Indoor Radar Detection: An FMCW Radar System for Distinguishing Human Presence and Swinging Blinds

This paper introduces an innovative indoor presence detection system with Frequency-Modulated Continuous-Wave (FMCW) radar, specifically engineered to address the challenge posed by swinging blinds in indoor environments. A prevalent issue in radar human detection for smart homes is the striking similarity between the motion of blinds swayed by natural wind and the moving patterns linked to human respiration. To overcome this challenge, our approach integrates motion frequency information using a Three-Dimensional Convolutional Neural Network (3D CNN) and utilizes deep metric learning to effectively distinguish between highly similar data from different classes. Remarkably, our model has achieved a classification accuracy of 96.85%, significantly surpassing the performance of existing state-of-the-art reconstruction-based methods. Additionally, our method is optimized for high operational efficiency, delivering real-time predictions every three seconds. This advancement highlights a contribution to the field of indoor radar presence detection, offering a more robust and efficient solution for smart home environments.