Using Skeleton based Optimized Residual Neural Network Architecture of Deep Learning for Human Fall Detection

For the last few decades, human posture recognition has gained mammoth popularity specifically in elderly fall detection. This study proposes a modified deep learning neural network architecture using OpenPose library for two category problem of human fall detection. OpenPose library computes and draws 18 points skeleton on frames. The proposed approach achieves an accuracy of 97.46% on publicly available, standard Multi Camera Fall dataset and outperforms existing state-of-the-art deep learning image classification Residual Network architectures ResNet-50 and ResNet-101 which have an accuracy of 94.63% and 70.18% respectively.