Hand Gesture Recognition Systems with the Wearable Myo Armband

Abstract

The hand gesture recognition systems deal with identifying a given gesture performed by the hand. This work addresses a hand gesture recognition method to classify and recognize the numbers from 0 to 9 in Turkish Sign Language based on surface electromyography (EMG) signals collected from a wearable device, namely the Myo armband. To accomplish such a goal, we have utilized machine learning techniques to recognize the hand gestures. In this context, seven different time domain features are extracted from the raw EMG signals using sliding window approach to get distinctive information. Then, the dimension of the feature matrix is reduced by using the principal component analysis to reduce the complexity of the deployed machine learning methods. The presented study includes the design, deployment and comparison of the machine learning algorithms that are k-nearest neighbor, support vector machines and artificial neural network. The results of the comparative comparison show that the support vector machines classifier based system results with the highest recognition rate.