Page 51 - IJEEE-2023-Vol19-ISSUE-1
P. 51

Received: 19 July 2022                 Revised: 28 August 2022  Accepted: 28 August 2022
DOI: 10.37917/ijeee.19.1.6
                                                                                          Vol. 19| Issue 1| June 2023

                                                                                          Ð Open Access

Iraqi Journal for Electrical and Electronic Engineering

Original Article

Wavelet-based Hybrid Learning Framework for
            Motor Imagery Classification

                      Z. T. Al-Qaysi1, Ali Al-Saegh*2, Ahmed Faeq Hussein3, M. A. Ahmed1
1 Department of Computer Science, Computer Science and Mathematics College, Tikrit University, Tikrit, Iraq

         2 Computer Engineering Department, College of Engineering, University of Mosul, Mosul, Iraq
 3 Biomedical Engineering Department, Faculty of Engineering, Al-Nahrain University, 10072, Baghdad, Iraq

Correspondence
*Ali Al-Saegh
Computer Engineering Department, College of Engineering,
University of Mosul, Mosul, Iraq
Email: ali.alsaegh@uomosul.edu.iq

Abstract
Due to their vital applications in many real-world situations, researchers are still presenting bunches of methods for better
analysis of motor imagery (MI) electroencephalograph (EEG) signals. However, in general, EEG signals are complex because
of their nonstationary and high-dimensionality properties. Therefore, high consideration needs to be taken in both feature
extraction and classification. In this paper, several hybrid classification models are built and their performance is compared.
Three famous wavelet mother functions are used for generating scalograms from the raw signals. The scalograms are used for
transfer learning of the well-known VGG-16 deep network. Then, one of six classifiers is used to determine the class of the
input signal. The performance of different combinations of mother functions and classifiers are compared on two MI EEG
datasets. Several evaluation metrics show that a model of VGG-16 feature extractor with a neural network classifier using the
Amor mother wavelet function has outperformed the results of state-of-the-art studies.
KEYWORDS: Brain-Computer Interface, Deep Learning, Motor Imagery, Transfer Learning, Wavelet Transformation.

                        I. INTRODUCTION                         wheelchairs and even self-driving cars [4]. Imagine moving
                                                                your body part without really moving that body part, which
    Technology for human-computer interaction has evolved       is known as MI [5]. EEG signals are generated by both
quickly in recent years, and the bioelectricity of the human    imagined and actual human movement. In motor imaging,
body is being developed as an interactive medium. A brain-      the EEG signals generated exhibit event-related
computer interface (BCI) system uses brain impulses to          synchronization (ERS) and event-related desynchronization
operate auxiliary equipment as a novel method of human-         (ERD) features [6]. There are four lobes in each of the human
computer interaction [1]. In a BCI system, a direct link        brain hemispheres, each serving a distinct purpose. Fissures
between the brain and a computer is established, bypassing      divide the lobes of the ear (sulcus). In the BCI system, the
the peripheral nervous system and providing a                   primary somatic sensory cortex (parietal lobe) and the
communication channel. ALS (Amyotrophic lateral                 primary motor cortex (temporal lobe) are the most critical
sclerosis), cerebral palsy, and motor neuron disease (MND)      areas [7]. Mu and beta rhythms in the sensorimotor part of
are all examples of brain illnesses that can benefit from BCI   one's hemisphere drop or increase as one imagines or
technology (MND) [2]. An EEG is the most preferred              performs the movement of a unilateral limb.
physiological sensor for developing a BCI system since it       Desynchronization caused by an event (ERD) and
meets both convenience criteria (i.e., non-intrusiveness and    synchronisation caused by an event (ERS) are two different
simplicity) as well as efficacy criteria (such as accuracy)     concepts [7].
(i.e., sensitivity, efficiency, and compatibility) [3]. P300
evoked potentials, steady-state visual evoked potentials            Fundamentally, MI-based BCI pattern recognition
(SSVEP), and Motor imagery (MI) are among the most              systems require three essential processes, namely
prominent EEG signal analysis disciplines [1]. Only MI          preprocessing of the EEG signal, feature extraction, and
relies on spontaneous potential and does not require any        classification [1]. Essentially, another crucial process in the
external stimulation. Researchers have employed MI signals      MI EEG pattern recognition model is the process of feature
to assist handicapped people in managing equipment like         extraction. Practically, extracted features are intended to

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and

reproduction in any medium, provided the original work is properly cited.

© 2022 The Authors. Published by Iraqi Journal for Electrical and Electronic Engineering by College of Engineering, University of Basrah.

https://doi.org/10.37917/ijeee.19.1.6                                                     https://www.ijeee.edu.iq 47
   46   47   48   49   50   51   52   53   54   55   56