Human Movement Recognition System Using CW Doppler Radar Sensor with FFT and Convolutional Neural Network

This paper presents a novel human movement recognition system using continuous-wave Doppler radar and deep learning with convolutional neural networks. The proposed monitoring system using Doppler sensor has environmental robustness, dark/light-independence, and are privacy-protected unlike traditional cameras. It works at a distance and does not cause discomfort to the elderly and patients.The proposed system was evaluated with various types of deep learning, including VGG-19, AlexNet, and GoogleNet CNNs (convolutional neural networks). The better performance of the monitoring system reaches 94.29% accuracy using VGG-19 CNN for detecting activities like falling, stopping, jumping, walking, and crouching. The Doppler radar sensor is inexpensive and an alternative to consider when the privacy of the elderly and patients is necessary. The proposed monitoring system of the elderly and patients can be used in an IoT (Internet of Things) network with multiple Doppler sensors scattered around the environment.