Feature Selection of EMG Signals Based on The Separability Matrix and Rough Set Theory

Feature Selection of EMG Signals Based on The Separability Matrix and Rough Set Theory

  This work considers the problem of selecting features of EMG signals in order to effectively use EMG signals to command the operations of mechatronic devices and systems. We first note that EMG signals form patterns that depend much on the subject from which the signals are extracted. We then introduce a notion called “separability matrix” in order to check the effectiveness of a given set of features for classification. Based on the separability matrix and in the form of the rough set theory, we present an algorithm of selecting features to minimize the subject-dependency. It is shown through an experimental study that the induced feature set obtained by the proposed feature selection algorithm has less subject-dependency than the other existing methods.  This work considers the problem of selecting features of EMG signals in order to effectively use EMG signals to command the operations of mechatronic devices and systems. We first note that EMG signals form patterns that depend much on the subject from which the signals are extracted. We then introduce a notion called “separability matrix” in order to check the effectiveness of a given set of features for classification. Based on the separability matrix and in the form of the rough set theory, we present an algorithm of selecting features to minimize the subject-dependency. It is shown through an experimental study that the induced feature set obtained by the proposed feature selection algorithm has less subject-dependency than the other existing methods.