ARTIFICIAL NEURAL NETWORK BASED ON ENSEMBLE EMPIRICAL MODE DECOMPOSITION FEATURES FOR HUMAN GAIT DIAGNOSIS AND CLASSIFICATION

Human gait analysis has been widely used to assess the stage of disease affecting the walking ability. The gait signals, namely vertical ground reaction force signals, become more unsteady and non-linear with the progress of the disease. This paper makes use of ground reaction force signals measured from both normal and subjects diagnosed with Parkinson. New features are then extracted from different intrinsic mode functions as a result of the ensemble mode decomposition. The extracted features are divided randomly into a training set of 60%, a validation set of 15% and testing set of 25%. The neural network is then employed which yield an interesting overall classification accuracy of 95.7 %. This paper will pave the way for better rehabilitative programs, understanding of gait biomechanics and fall prevention among the elderly.