Recognition of Unbalanced Human Gait by Using RNN-LSTM Algorithm

Abstract

This study proposes a method to recognize unbalanced human gait by using Recurrent neural networks - long short term memory models (RNN-LSTM) algorithm with 8 Kinect sensors. The obtained result by the developed multiple Kinect system is higher than at least 93% accuracy when compared to Optitrack. In addition, we obtained 98.88 % test accuracy and 0.0554 loss for recognizing the abnormal gait by deep learning training RNN-LSTM algorithm.