Classification of EEG Mental Tasks using Multi-Objective Flower Pollination Algorithm for Person Identification

Abstract

In the modern life, the authentication technique for any system is considered as one of the most important challenges task which must careful consideration. Therefore, many researchers have developed traditional authentication systems to deal with our digital world. Recently, The Biometric techniques have been successfully provided a high level of authentication, such as fingerprint, face recognition, and voice recognition. In this paper, a new authentication system has been proposed which is based on EEG signals with hybridizing wavelet transform and multi-objective flower pollination algorithm (MOFPA-WT). The main task of MOFPA is to find the optimal WT parameters for EEG signal denoising which can extract unique features form the EEG. The proposed method (MOFPA-WT) tested using a standard EEG database which has five different mental tasks, includes baseline, multiplication, rotation, letter composing, and visual counting. To classify the EEG signals using proposed method four classification methods are applied which are, neural network, decision tree, Naive Bayes, and support vector machine. The performance of the (MOFPA-WT) is evaluated using four criteria: (i) accuracy, (ii) sensitivity, (iii) specificity, (v) false acceptance rate. The experimental results show the (MOFPA-WT) can achieve the highest recognition rates up to 85% using neural network classifier based on visual counting task as well as the EEG(_Std) feature obtained the highest accuracy compared with others EEG features based on visual counting task.