Centroid-Based Security Monitoring for Indoor Elderly Care with Millimeter Wave Radar

The increasing use of millimeter wave radar for smart elderly care is due to its precision, privacy, and cost-effectiveness. This technology enables real-time monitoring of unexpected incidents, particularly falls among the elderly indoors. By examining the relationship between cloud points and human centroids, this paper exploits the unique features of millimeter wave radar for personnel detection and human activity recognition. An improved Kalman filter algorithm for centroid tracking has been introduced. It effectively monitors the three-dimensional trajectory and velocity changes associated with various human behaviors. Subsequently, machine learning algorithms and neural networks are employed to classify these behaviors, including walking, sitting, and falling. Compared to conventional Kalman filtering, this approach effectively handles dispersion characteristics within point clouds, resulting in a smoother trajectory and more consistent centroid. This method enables the recognition of abnormal activities of people indoors with a maximum classification accuracy of 93.02%.