| Abstract | In past years, researchers came up with the number of artifact removal techniques for elim inating artifacts in EEG signals ranging from traditional techniques like Wavelet transform, the techniques that overcame techniques’ limitations, wICA, and techniques that based on neural network, WNN. However, these invented techniques cannot be compared across stud ies since most of their performance metrics are distinct. Moreover, these techniques did not implemented in EEG applications resulting in the unclear relation between artifact removal techniques and the improvement in EEG application remains unclear. Therefore, in this study, we first, intrinsic evaluate 5 EOG artifact removal techniques (Wavelet transform, ICA, wICA, WNN and U-NET) and 2 ECG artifact removal techniques (ICA-correlation, ICA-ctps) on different performance metrics such as RMSE and Signal-to-Artifacts Ratio. The result show that U-NET outperform the other techniques for EOG artifact removal tech niques as a technique based on neural network while wICA also has a great performance. For ECG removal technique, ICA-ctps can capture most of R-peaks responded to ECG refer ence channel. Second, for extrinsic analysis, we implement these three techniques on N170 dataset and Motor imagery dataset. In conclusion, we should apply artifact removal tech niques on trials or epochs that are contaminated by artifact like EOG signals only to prevent the techniques distort epochs that did not contain any significant artifacts. |