Classification of motor imagery brain wave for bionic hand movement using multilayer perceptron

Abstract

Physical disability due to amputation can affect a person's quality of life due to limited movement in performing daily activities. Bionic hands are used to help someone with an amputation disability. This research developed a bionic hand control based on electroencephalography sensors capable of measuring the brain's bioelectric activity. The classified brain wave was then translated as activity pattern information. The alpha & beta waves were the focus of this work. This study demonstrated a method to extract and classify motor imagery of brainwave activity patterns. The Fast Fourier Transform (FFT) method extracts motor imagery characteristics. The extraction of features is then classified by the Multilayer Perceptron (MLP) method for five classes of bionic hand movement. Testing was conducted with two scenarios. The first test motor imagery without additional movement showed an accuracy of 77.20 %, while the second test motor imagery combined with head movement showed an accuracy of 84.40% for five classes. The system based on motor imagery has been implemented in a bionic hand that shows the applicability of the proposed method.