Knee joint moment estimation using neural network system identification in sit-to-stand movement

In several studies, neural network was used to identify the relationship between EMG signals and joint moment. But those studies were mostly preformed in isokinetic movement with general feed forward neural network. In this study, we used NNARX(Neural Network, AutoRegressive, eXternal input) model structure which is one of identification model structure for nonlinear dynamic system to identify relationship between EMG signals and knee joint moment in sit-to-stand movement which is representative dynamic movement in daily human living. And validation of our proposed method was performed with simultaneously measured EMG signals and kinematic data during sit-to-stand movement. To compare with results of our method, identification using back-propagation neural network structure was also performed.