A convolution neural network approach for fall detection based on adaptive channel selection of UWB radar signals

Abstract According to the World Health Organization and other authorities, falls are one of main causes accidental injuries among elderly population. Therefore, it is essential detect predict fall activities older persons in indoor environments such as homes, nursing, senior residential centers, care facilities. Due non-contact signal confidentiality characteristics, radar equipment widely used care, detection, rescue. This paper proposes an adaptive channel selection algorithm separate activity signals from background using ultra-wideband generalize fused features frequency- time-domain images which will be sent a lightweight convolutional neural network recognize activities. The experimental results show that method able distinguish three types (i.e., stand fall, bow squat fall) obtain high recognition accuracy up 95.7%.