PROJECT TITLE :

Towards Zero Retraining for Myoelectric Control Based on Common Model Component Analysis

ABSTRACT:

In spite of several decades of intense analysis and development, the existing algorithms of myoelectric pattern recognition (MPR) are nonetheless to satisfy the factors that a sensible upper extremity prostheses ought to fulfill. This study focuses on the criterion of the short, or perhaps zero subject training. Thanks to the inherent nonstationarity in surface electromyography (sEMG) signals, current myoelectric control algorithms typically need to be retrained daily during a multiple days' usage. This study was conducted based on the hypothesis that there exist some invariant characteristics in the sEMG signals when a theme performs the identical motion in numerous days. Thus, given a group of classifiers (models) trained on several days, it's potential to find common characteristics among them. To this end, we tend to proposed to use common model component analysis (CMCA) framework, in that an optimized projection was found to minimize the dissimilarity among multiple models of linear discriminant analysis (LDA) trained using knowledge from totally different days. Five intact-limbed subjects and 2 transradial amputee subjects participated in an experiment as well as six sessions of sEMG data recording, which were performed in six different days, to simulate the application of MPR over multiple days. The results demonstrate that CMCA contains a vital higher generalization ability with unseen data (not included in the coaching data), resulting in classification accuracy improvement and increase of completion rate in a very motion take a look at simulation, when comparing with the baseline reference technique. The results indicate that CMCA holds a nice potential in the effort of developing zero retraining of MPR.