Electroencephalogram (EEG) artifacts such as eyeblink, eye movement, and muscle movements widely contaminate the EEG signals. Those unwanted artifacts corrupt the information contained in the EEG signals and degrade the performance of qualitative analysis of clinical applications and as well as EEG-based brain–computer interfaces (BCIs). The applications of wavelet transform in denoising EEG signals are increasing day by day due to its capability of handling non-stationary signals. All the reported wavelet denoising techniques for EEG signals are surveyed in this paper in terms of the quality of noise removal and retrieving important information. In order to evaluate the performance of wavelet denoising techniques for EEG signals and to express the quality of reconstruction, the techniques were evaluated based on the results shown in the respective literature. We also compare certain features in the evaluation of the wavelet denoising techniques, such as the requirement of reference channel, automation, online, and performance on a single channel.

This paper proposes an automatic eyeblink artifacts removal method from corrupted-EEG signals using discrete wavelet transform (DWT) and meta-heuristically optimized threshold. The novel idea of thresholding approximation-coefficients (ACs) instead of detail-coefficients (DCs) of DWT of EEG in a backward manner is proposed for the first time for the removal of eyeblink artifacts. EEG is very sensitive and easily gets affected by eyeblink artifacts. First, the eyeblink corrupted EEG signals are identified using support vector machine (SVM) as a classifier. Then the corrupted EEG signal is decomposed using DWT up to the sixth level. Both the mother wavelet and the level of decomposition are selected using appropriate techniques. Then the ACs are thresholded in backward manner using the optimum threshold values followed by inverse DWT operation to reconstruct the original EEG signal. The AC at level 6 is thresholded and is used in IDWT with DC to get back the AC at level 5. Likewise, the backward thresholding of the ACs followed by IDWT is continued till the artifact free EEG signal is reconstructed at level 1. The optimum values of the thresholds of the ACs at different levels are optimized using two meta-heuristic algorithms, particle swarm optimization (PSO) and grey wolf optimization (GWO) for comparison. The results reveal that the proposed methodology is superior to the recently reported methods in terms of average correlation coefficient (CC) which states that the proposed method is better in terms of the quality of reconstruction in addition to being fully automatic.

Electroencephalogram (EEG) is often corrupted with artifacts originating from sources like the eyes and muscles. Hybrid artifact removal methods often require human intervention for the adjustment of different parameters. We propose a robust method that can automatically detect and remove eye-blink and muscular artifacts from EEG using k-nearest neighbor (kNN) classifier and a long short-term memory (LSTM) network. Our method adopts a sliding window of 0.5 s to detect and remove the artifacts from EEG. Features, such as the variance, peak-peak amplitude, and average rectified value are calculated for each EEG segment to identify corrupted segments using the kNN classifier. The kNN classifier detects the presence of artifacts, after which, the corresponding EEG window is forwarded to the LSTM network for artifact removal. The LSTM network is trained with the corrupted segments of 0.5 s as input and clean segments of 0.5 s as output. Our method achieved an accuracy of 97.4% in identifying corrupted EEG segments and an average correlation coefficient, structural similarity, signal to artifact ratio, and normalized mean squared error of 0.69, 0.76, 1.52 dB, and 0.0013, respectively, in cleaning the EEG. Our results outperformed other hybrid methods reported in the literature based on combination of ensemble empirical mode decomposition and canonical correlation analysis, combination of independent component analysis and wavelet decomposition, and tensor decomposition. The mean absolute error of our method is also better in comparison to other methods. Our method can be applied to single and multiple channels and does not require any tuning of parameters.

Emotion recognition using EEG has been widely studied to address the challenges associated with affective computing. Using manual feature extraction methods on EEG signals results in sub-optimal performance by the learning models. With the advancements in deep learning as a tool for automated feature engineering, in this work, a hybrid of manual and automatic feature extraction methods has been proposed. The asymmetry in different brain regions is captured in a 2D vector, termed the AsMap, from the differential entropy features of EEG signals. These AsMaps are then used to extract features automatically using a convolutional neural network model. The proposed feature extraction method has been compared with differential entropy and other feature extraction methods such as relative asymmetry, differential asymmetry and differential caudality. Experiments are conducted using the SJTU emotion EEG dataset and the DEAP dataset on different classification problems based on the number of classes. Results obtained indicate that the proposed method of feature extraction results in higher classification accuracy, outperforming the other feature extraction methods. The highest classification accuracy of 97.10% is achieved on a three-class classification problem using the SJTU emotion EEG dataset. Further, this work has also assessed the impact of window size on classification accuracy.

This paper presents a collection of electroencephalogram (EEG) data recorded from 40 subjects (female: 14, male: 26, mean age: 21.5 years). The dataset was recorded from the subjects while performing various tasks such as Stroop color-word test, solving arithmetic questions, identification of symmetric mirror images, and a state of relaxation. The experiment was primarily conducted to monitor the short-term stress elicited in an individual while performing the aforementioned cognitive tasks. The individual tasks were carried out for 25 s and were repeated to record three trials. The EEG was recorded using a 32-channel Emotiv Epoc Flex gel kit. The EEG data were then segmented into non-overlapping epochs of 25 s depending on the various tasks performed by the subjects. The EEG data were further processed to remove the baseline drifts by subtracting the average trend obtained using the Savitzky-Golay filter. Furthermore, the artifacts were also removed from the EEG data by applying wavelet thresholding. The dataset proposed in this paper can aid and support the research activities in the field of brain-computer interface and can also be used in the identification of patterns in the EEG data elicited due to stress.

Emotion plays a vital role in understanding the affective state of mind of an individual. In recent years, emotion classification using electroencephalogram (EEG) has emerged as a key element of affective computing. Many researchers have prepared datasets, such as DEAP and SEED, containing EEG signals captured by the elicitation of emotion using audio–visual stimuli, and many studies have been conducted to classify emotions using these datasets. However, baseline power removal is still considered one of the trivial aspects of preprocessing in feature extraction. The most common technique that prevails is subtracting the baseline power from the trial EEG power. In this paper, a novel method called InvBase method is proposed for removing baseline power before extracting features that remain invariant irrespective of the subject. The features extracted from the baseline removed EEG data are then used for classification of two classes of emotion, i.e., valence and arousal. The proposed scheme is compared with subtractive and no-baseline-correction methods. In terms of classification accuracy, it outperforms the existing state-of-art methods in both valence and arousal classification. The InvBase method plus multilayer perceptron shows an improvement of 29% over the no-baseline-correction method and 15% over the subtractive method.