Proposed Model for Thought-Based Animation based on Classifying EEG signals using Estimated Parameters and Multi-SVM

Abstract

Brain Computer Interface (BCI) is a powerful tool to assist people. In this paper we work on interpreting motor imagery tasks. We propose a model based on estimating statistical parameters of the Electroencephalography (EEG) signal and using these as features. We then feed the features vector to a multi-class Support Vector Machine (SVM) for classification. Promising results were obtained by testing the proposed model on the publicly available BCI competition 2008 dataset. An average classification rate of 90.2% and a kappa result of 0.86 were achieved. The kappa result is considered a very good agreement. We further show an application for animating characters using the classification output from the EEG signals.