Radar Based Real-Time Fall Detection System with Low Power Consumption

Fall detection plays a critical role in the elder people home caring. In this work, we propose a low-power fall detection system based on Doppler radar sensor. To lower the power consumption which is of great importance in practical applications, the system is designed to be multistage including event detection, fall-like event detection and fall detection parts. By adopting the proposed multi-stage system, the most power consumption part, deep neural network operation, will not be activated in most cases. The experiment results show that the proposed model can not only achieve high accuracy of fall detection, but also have great potential for deployment in a low power mode.