Implementation of Fall Detection for The Elderly Home Support using 24-GHz Radar Sensor

In this paper, we propose a development of a smart home system for assisting elderly people by implementing a fall detection using 24-GHz radar sensor. The system is mounted on the ceiling of the room and the system records the doppler frequency shift signals of a body moving with a duration of 2.5 seconds. The recorded signals are transformed into spectrogram images with a size of 119 x 65 pixels. Then the spectrogram images are recognized by the Convolutional Neural Network (CNN) algorithm. There are 3 classes to be recognized, namely: falling, vacant, and walking. The dataset was acquired, built, and trained by using Edge Impulse platform. The hardware used in the deployment phase is Arduino Nano 33 BLE Sense. The experimental results show an accuracy rate of the training phase and the testing phase of 98.1% and 96.30%, respectively.