Recognition of phases in sit-to-stand motion by Neural Network Ensemble (NNE) for power assist robot

Power assist robot (PAR) would be of great value to certain human activities or for special groups such as the aged and the weakling. If PAR can provide appropriate power assist, it would be more effective to adapt to human intention accordingly, otherwise, it would be a burden to human body, especially for those have difficulty in standing up and sitting down. In this paper, in order to realize proper power assist during human´s standing up, a method for recognizing successive phases during a subject performing sit-to-stand motion is proposed. By using of surface electromyographic (EMG) signal sampled from lower limb combined with floor reaction force (FRF) which indicates the position of centre of gravity (COG) of some a subject, sufficient information for sit-to-stand could be recorded when the subject stands up from the chair. In succession, characteristics of the recorded signals are extracted. As it is expected for high recognition rate and rather robustness method, neural network ensemble (NNE, hereafter) technology is selected to identify each phase included in standing up.