Decreasing Parameters of Deep Neural Network Model for Fall Detection
The goal of this work is to achieve a small model which have huge possibilities to implements on small
embedded system. We use convolutional neural network (CNN) to classify human’s 3D skeleton information as falling
cases. After deploying several simple rules, the network is small enough to execute without CUDA support platform and
retains the accuracy over 99% on NTURGB+D dataset.
Keywords - Deep Neural Network, Fall Detection, Public Health Problem, Artifactual Intelligence