Deep Convolution Neural Network and Random Forest Algorithm for BCI based Photo Imagery Learning

In a brain computer interface (BCI) based system, the motor imagery (MI) based classification of electrocorticograms (ECoGs) is proposed in this paper with deep neural network model. When compared to the traditional classification and feature extraction models, the proposed model offers improved performance. Classification is performed using traditional algorithm while feature extraction is performed using the deep learning algorithm in the proposed model. The data is trained and feature extraction is performed using deep convolution neural network (DCNN) and combined with Random Forest (RF) algorithm for classification of features. The brain activities are observed and feature information is obtained using the RF and DCNN algorithms. The human body action is used for obtaining classification results. BCI competition dataset is used for the performance evaluation of the proposed framework. This work opens new opportunities for BCI system based future research avenues with the combination of traditional algorithms and deep learning models.