Page 59 - IJEEE-2023-Vol19-ISSUE-1

Al-Qaysi1, Al-Saegh, Hussein & Ahmed | 55

TABLE III
WTNN EVALUATION OVER DATASET-II (EVALUATION PART)

Performance Metrics

Subjects Training Testing AUC CA F1 Precision Recall Logloss Specificity
Time Time
S1 381.68 17.501 0.999 0.997 0.997 0.997 0.997 0.020 0.997
S2 293.11 15.572 1.00 0.997 0.997 0.997 0.997 0.008 0.997
S3 317.75 18.519 0.992 0.994 0.994 0.994 0.994 0.129 0.994
S4 83.14 4.244 1.00 1.00 1.00 1.00 1.00 0.006 1.00
S5 118.32 4.601 1.00 0.998 0.998 0.998 0.998 0.014 0.998
S6 104.40 6.487 1.00 0.997 0.997 0.997 0.997 0.011 0.997
S7 128.03 7.797 1.00 0.994 0.994 0.994 0.994 0.011 0.994
S8 123.33 6.893 0.990 0.994 0.994 0.994 0.994 0.048 0.994
S9 117.84 6.605 1.00 0.994 0.994 0.994 0.994 0.015 0.994
Mean 185.28 9.802 0.998 0.99 0.996 0.996 0.996 0.029 0.996

TABLE IV

RESULTS COMPARISON WITH STATE-OF-THE-ART STUDIES [15] D.-K. Kim, et al., "Sequential Transfer Learning via
Segment After Cue Enhances the Motor Imagery-based
RELATED TO DATASET-I Brain-Computer Interface," in 2021 9th International
Winter Conference on Brain-Computer Interface (BCI),
Year Study Method Accuracy 2021, pp. 1-5.

2015 [35] LDA + based wrapper SFS 90% [16] K.-T. Kim, et al., "Subject-Transfer Approach based on
Convolutional Neural Network for the SSSEP-BCIs," in
2016 [36] STFT with KNN 83.57% 2021 9th International Winter Conference on Brain-
Computer Interface (BCI), 2021, pp. 1-3.
2016 [37] WT + SE using SVM and KNN 86.4%
[17] N. K. Al-Qazzaz, et al., "Selection of mother wavelet
2016 [38] MEMD + STFT with KNN 90.71% functions for multi-channel EEG signal analysis during a
working memory task," Sensors, vol. 15, pp. 29015-29035,
2017 [39] Fuzzi?ed Adaptation with SVM 81.48% 2015.

2019 [40] Genetic Algorithm with FKNN 84% [18] N. K. Al-Qazzaz, et al., "EEG Feature Fusion for Motor
Imagery: A New Robust Framework Towards Stroke
2019 [41] STFT with CNN 89.73% Patients Rehabilitation," Computers in biology and
medicine, p. 104799, 2021.
2019 [42] CWT with 1D CNN 92.9%
[19] G. Xu, et al., "A deep transfer convolutional neural
2020 [30] WPT + CWT with CNN 95.71% network framework for EEG signal classification," IEEE
Access, vol. 7, pp. 112767-112776, 2019.
2021 [43] WTTD + CWT with CNN 96.43%
[20] R. Mahamune and S. H. Laskar, "Classification of the
2022 This CWT (Amor and Morlet) + 100% four-class motor imagery signals using continuous wavelet
Study VGG-16 + NN transform filter bank-based two-dimensional images,"
International Journal of Imaging Systems and Technology,
TABLE V 2021.

RESULTS COMPARISON WITH STATE-OF-THE-ART STUDIES [21] A. M. Álvarez-Meza, et al., "Time-series
discrimination using feature relevance analysis in motor
RELATED TO DATASET-II imagery classification," Neurocomputing, vol. 151, pp.
122-129, 2015.
Year Study Method Accuracy
[22] J. Wang, et al., "Toward optimal feature and time
2014 [24] Hjorth parameter + LDA 79.1% segment selection by divergence method for EEG signals
classification," Computers in biology and medicine, vol.
2015 [21] CSP + EMD 72.30% 97, pp. 161-170, 2018.

2018 [22] CSP + autoregressive model 77% [23] C. Xu, et al., "Two-level multi-domain feature
extraction on sparse representation for motor imagery
2018 [28] WDPSD 89.36% classification," Biomedical Signal Processing and
Control, vol. 62, p. 102160, 2020.
2018 [29] A normalization model with one 96.86%
contralateral EOG channel [24] S.-H. Oh, et al., "A novel EEG feature extraction
method using Hjorth parameter," International Journal of
2019 [26] a separated channel 83%
convolutional network

2019 [19] STFT + VGG16 71.2%

2020 [23] multi-domain features 79%

2020 [27] CNN with hybrid convolution 87.6%
scale

2020 [25] Hilbert transform (HT)-SVM 82.50%

2020 [30] CWT + VGG19 97.06%

2021 [20] CWT + CNN 71.25

2022 This CWT (Amor and Morlet) + 99%
Study VGG-16 + NN