Fall Detection with Neural Networks
Gaojing Wang, Zheng Liu, Qingquan Li · 2019
Falls have been one of the main threats to people's health, especially for the elderly. Detecting falls in time can minimize the severity of injury and save lives. On the other hand, neural networks have shown superior performance in human activity recognition field, but their performance in detecting falls has barely been evaluated. In this paper, benefits of using neural networks, including feed-forward, recurrent and convolutional neural networks, in fall detection based on wearable sensors were explored. Experiment results showed that neural networks performed much better than conventional machine learning methods. The best result was obtained from the convolutional neural network with a sensitivity of 99.05% and a specificity of 99.68%.