KNN Based GA for Performance Improvement in Neck Movement Classification of EMG signal

Providing significant features to the classifier is essential for the enhancement of classification accuracy. To meet this demand, feature selection plays a vital role in the field of bio signals. This work engaged in the classification of movements based on electromyography (EMG) signals acquired from the selected neck muscles namely the trapezius and sternocleidomastoid. For the relevant feature selection, an optimization technique named genetic algorithm (GA) is implemented. This GA is an iterative process that is involved in performance improvement by reducing the classification error. Here, the k-nearest neighbor (KNN) algorithm is used iteratively from which the error function is returned as a fitness function that is used to fit the selected individual to an optimal solution. The selected features are classified by the application of KNN, support vector machine (SVM), and Naive Bayes (NB) classifier, of which KNN classifier defeats others with a better performance of about 99.93±0.14%.