American Sign Language Recognition using Electromyographic Signals via Myo Armband

Abstract

Human and computer interaction has become an important tool in the 21 st century One of the most promising technologies for human/computer interaction is speech recognition ( which is a system composed of microphones, soundcards and speech engine software in which isolate words, spoken by an emitter is recognized through calculation of a minimum prediction residual (Johnson et al 2014 Nowadays there has been many studies around this form of technology but the hearing impaired population cannot benefit from these advancements This study proposes to address this problem by applying the Myo Armband, a wearable commercial band that captures Electromyographic signals ( and tests its potential to correctly classify American Sign Language ( gestures by applying a brain computer interface methodology in which consist of three parts data acquisition, signal processing and classification This methodology was applied to three types of hand gestures with the objective of extracting a mean power value for each sensor and compare it to experimental signals to determine if a correct classification was obtained In the end, the brain computer interface methodology was accomplished successfully, as it was possible to measure the power of every sensor After finalizing the research, it is possible to say that the approach taken to satisfy the primary goal, which was recognizing the three gestures, was not accurate, as it did not reach a satisfying accuracy percentage There can be many factors of why this happened One can be the fact that the algorithm created and the approach of measuring the power of every sensor of the Myo Armband was not the indicated or the best to recognize EMG signals or that the sample experimental size was very small One potential candidate to enhance the obtained results is to increase the number of experimental samples or to develop a new approach to process the obtained signals and enhance the classification