Evaluating a biosensor-based interface to recognize hand-finger gestures using a Myo armband

Abstract

Gesture recognition is a convenient and natural Human-Computer Interaction (HCI) technique. Recent advances in bioengineering have seen the use of biosensor technologies in HCI, since various biosensors provide real-time feedback from biological activities. This has enabled User Interface (UI) designers to design more natural UIs, including the use of a Muscle-Computer Interface (MCI). This paper presents an evaluation of an MCI designed to recognize four-finger pinching, fist action and five-hand-fingers spread gestures using Electromyography (EMG) signals from a Myo armband from Thalmic Labs Inc.™. An experimental research strategy was used and a Feedforward Neural Network was implemented to classify the gestures and each of the gestures was trained for 3 seconds.An average 95% success rate for completing gesture-posing tasks among 6 participants was achieved with an average predicting error value of 14.48, expressed by the Root Mean Square Error (RMSE). The results illustrate the application of biosensors in gesture recognition as a modern and reliable approach which benefits the HCI. Biosensor-based gesture recognition provided a greater level of accessibility and encumbered the users less when this approach was compared with vision- and sensor-based approaches. The Myo armband showed that it is an economical and standard bio-sensing wearable device that can be successfully used in hand-finger gesture recognition.