Real-Time Fall Detection Using Microwave Doppler Sensor-Computational Cost Reduction Method Based on Genetic Algorithm

The risk of falling is a serious problem for the elderly. If the fall remains undetected for a considerable period, hypothermia and dehydration may occur, leading to death. Therefore, real-time detection of falls is critical. We previously proposed a fall-detection system based on a microwave Doppler sensor, which can capture the object's velocity. Our system performs template matching based on the dynamic time warping distance; hence, the processing time depends on the number of template datasets. Herein, we attempt a real-time fall detection by reducing the number of templates. We apply the genetic algorithm to select better templates. The fitness of the selected templates is evaluated by dividing the accuracy by the number of templates.