Sign Translation with Myo Armbands

Abstract

© 2017 IEEE. Sign language is a common non-verbal communication method for people with impaired hearing. Despite the existence of sign language, there are communications boundaries. With the implementation of Myo armbands in predicting the gestures, this gap could be reduced. Myo armbands can collect three-axis accelerometer, three-axis gyroscope, and three-axis magnetometer signals. These collected data are used in constructing the prediction model. Translation is possible with the use of machine learning algorithms implemented in the prediction model. The algorithms used for testing the best accuracy rate from the prediction model are Decision Tree, Sequential Minimal Optimization, and Multilayer Perceptron. In addition, Mean and Standard Deviation (SD) are used to test for optimal feature selection. After testing, the prediction model that gave the best result were with MLP and SMO algorithms having Mean and SD as the optimal features. Thus, either of them could be implemented.