Efficient energy smart sensor for fall detection based on accelerometer data and CNN model

Fall detection helps to provide medical assistance quickly and to avoid the aggravation of injuries. In this paper, we propose a new noninvasive and energy-efficient smart sensor for fall detection. The sensor is based on accelerometer data and is attached to 20 building workers. To reduce power consumption, a new method of data selection is proposed. This method is based on the use of sensor timers, which allows for the reduction of 91% of the acquired data and 94% of the transmitted data. Regarding the classification, a new classification approach is proposed. Indeed, each data segment is displayed as a graph. Then, a convolution neural network is trained to detect the presence or absence of falls in each graph. An accuracy of 98% was obtained. This result exceeds that obtained in several studies and shows the effectiveness of the proposed approach.