Fall detection based on LCNN and fusion model of weights using human skeleton and optical flow.

Falls are the main cause of increased morbidity and mortality in the elderly. Human activity recognition is a difficult task due to factors such as light changes and complex background. To address these challenges, this paper proposes a method for recognizing falling activity based on the human skeleton and optical flow sequence. First, videos of fall activities and not-fall activities of different testers are captured using the Kinect v2 sensor; then, the captured videos are pre-processed to obtain the skeleton images and grayscale images of each activity; and the optical flow sequences of adjacent images are obtained using the dense optical flow algorithm. This paper proposes a lightweight convolutional neural network for extracting movement features from multimodal data and identifying activities. The decision fusion algorithm combines the fuzzy comprehensive evaluation method and the majority voting strategy to determine the weight values that influence the final decision outcomes. The experimental results show that the accuracy of the method proposed in this paper is up to 95.31%, which has a good effect on the recognition of falling activity. [ABSTRACT FROM AUTHOR]