Classification of Motor Imaginary Signals Using Stacked Autoencoders

Motor imagery signals are one of the electroencephalography signals that occurs by subjects imagining the movement of a body limb and they used in brain computer interface applications. With the use of deep neural networks in motor imagery signal classification, classification performance has increased. Deep belief networks are one of the networks that increase performance. In this study, using a stacked autoencoders and BCI IV 2b dataset, a deep belief network and education method with more robust and higher performance is proposed. Average classification performance is obtained as 80.44%.