Motor imagery EEG signal classification using image processing technique over GoogLeNet deep learning algorithm for controlling the robot manipulator

Abstract

Controlling of a robotic arm using a brain-computer interface (BCI) is one of the most impressive applications. In this study, a novel method for the classification of motor imaging (MI) electroencephalography (EEG) signals are proposed for BCI. EEG signals are divided into three secondary tables, which were converted into spectrogram images. After applying the spectrogram method, the obtained images are divided into folder structures and deep learning is performed. In the deep learning stage, 400 images are obtained for each task as input to the Goo-gLeNet. After the deep learning completed, the presented system has been tested to imagine up, down, left and right movement to control the movement of the robot arm. It is observed that the robot arm performs the desired movement over 90% accuracy.