Evaluating the Feasibility of EMG and Bend Sensors for Classifying Hand Gestures

Authors: 
R. Tidwell, S. Akumall, S. Karlaputi, R. Akl, K. Kavi and D. Struble
Keywords: 
Surface EMG sensors, Bend resistive sensors, SVM model, Accelerometers.
Abstract: 

This paper presents a feasibility study using Surface Electromyography (EMG) and bend resistive sensors for hand gesture recognition. Using only surface EMG signals, we classified gestures using a neural network, logistic regression, and a Support Vector Machine (SVM) model. The results ranged widely from 40-75% accuracy. To improve accuracy, we included bend resistive sensors that are attached to each of the five fingers of the hand. This approach is sometimes known as a “data glove” and is an alternative to using surface EMG only for gesture recognition. Using the data glove accompanied with an SVM model, we achieved 93.33% accuracy over a 10 hand gesture set. With these sensors, we are able to classify hand gestures, and not arm movements. Future work will be incorporating accelerometers and varied placement of EMG sensors into the model to widen the range of gesture recognition and increase accuracy.

Publish Date: 
Thursday, July 18, 2013
Venue: 
Proceedings of the International Conference on Multimedia and Human Computer Interaction (MHCI-13)
Paper URL: 
http://csrl.cse.unt.edu/~kavi/Research/MHCI-13.pdf