Electroencephalography (EEG) is invariably contaminated by artifacts, which can delay with the underlying neural oscillatory patterns and impair the accuracy of subsequent analyses. In this research, an automated technique is proposed for artifact removal (AR) in EEG signals based on wavelet transform with a threshold value. The research focuses on the use of parameterization of the mother wavelet to optimize the AR without any significant effect on the signal. The EEG data is collected from 40 students and the original recording is set; then, the data is down-sampled to 128 Hz. Optimal mother wavelet (OMW)-based AR is proposed in this research, which is a wavelet-based denoising technique that is used by varying threshold in a specific interval. Then, a non-negative garrote shrinkage function is applied for denoising which can take relevant features and generate a higher AR level with the lowest mean square error (MSE). Finally, the inverse stationary wavelet transform (ISWT) is applied to obtain an artifact-reduced EEG sequence. This approach is validated by extracting of several statistical features like kurtosis, variance, range, and Shannon’s entropy from the independent components (ICs) of the EEG signals. The experimental results showed that the proposed OMW-AR method achieved better accuracy of 98.95% in removal of artifacts from EEG and lower MSE of 4.21.