Extracted From The Electroencephalogram Research Articles

Deep Learning Method for Selecting Effective Models and Feature Groups in Emotion Recognition Using an Asian Multimodal Database

Emotional awareness is vital for advanced interactions between humans and computer systems. This paper introduces a new multimodal dataset called MERTI-Apps based on Asian physiological signals and proposes a genetic algorithm (GA)—long short-term memory (LSTM) deep learning model to derive the active feature groups for emotion recognition. This study developed an annotation labeling program for observers to tag the emotions of subjects by their arousal and valence during dataset creation. In the learning phase, a GA was used to select effective LSTM model parameters and determine the active feature group from 37 features and 25 brain lateralization features extracted from the electroencephalogram (EEG) time, frequency, and time–frequency domains. The proposed model achieved a root-mean-square error (RMSE) of 0.0156 in terms of the valence regression performance in the MAHNOB-HCI dataset, and RMSE performances of 0.0579 and 0.0287 in terms of valence and arousal regression performance, and 65.7% and 88.3% in terms of valence and arousal accuracy in the in-house MERTI-Apps dataset, which uses Asian-population-specific 12-channel EEG data and adds an additional brain lateralization (BL) feature. The results revealed 91.3% and 94.8% accuracy in the valence and arousal domain in the DEAP dataset owing to the effective model selection of a GA.

Dimensions of anxiety and depression and neurophysiological indicators of error-monitoring: Relationship with delta and theta oscillatory power and error-related negativity amplitude.

Error-monitoring processes may be affected by transdiagnostic dimensions of psychopathology symptoms including trait anxiety, worry, and severity of depressive symptoms. We tested the relationship between continuous measures of anxiety and depressive symptomology and neural correlates of error-monitoring as measured by time-frequency domain delta and theta oscillatory power and time-domain error-related negativity (ERN) amplitude extracted from the electroencephalogram (EEG). Secondary analyses tested for diagnostic group differences in error-related neural responses in individuals with generalized anxiety disorder (GAD), major depressive disorder (MDD), and comorbid psychiatric disorders. About 178 participants (104 female, M[SD]age =21.7[4.6]) with a wide range of psychopathology symptoms completed a modified version of the Eriksen flanker task and symptom questionnaires. Residualized difference values between correct and error trials for delta/theta power and error/correct ERN amplitude were used as dependent variables. Linear regression analyses adjusted for age, sex, and task accuracy showed nonsignificant associations of symptom dimension measures with error-related residualized delta/theta power or residualized ERN amplitude. Subset analyses on those with confirmed psychopathology diagnoses also did not predict residualized error-related delta/theta power nor residualized ERN amplitude (nGAD =14, nMDD =28, nComorbid =19, nControl =85). Taken in the context of the previous literature, results suggest a heterogeneous relationship between depressive and anxiety symptom dimensions and neurophysiological indices of error-monitoring.

Classification of EEG signals using hybrid combination of features for lie detection

The present work demonstrates the effectiveness of the combination of time, frequency, time–frequency, and statistical features extracted from the electroencephalogram (EEG) data, with support vector machine (SVM) for lie detection. Predominantly, the features extracted from the empirical mode decomposition (EMD) of the EEG data significantly improve the classification accuracy. A specific number of narrow band oscillatory components, called intrinsic mode functions (IMFs), are obtained after EMD of the data. The first three IMFs are selected to extract three time and three frequency domain statistical features corresponding to each IMF. These features are chosen due to the strong data adaptation capability of EMD for the transient signals such as an EEG. Furthermore, the features are selected keeping in mind the differences in the distribution, average value, and regularity of the guilty and innocent subjects’ brain signals. The proposed combination of extracted features with customized SVM demonstrates better accuracy than the other state-of-the-art feature extraction methods reported earlier. The proposed hybrid combination of features prominently distinguishes the guilty and innocent subjects with the classification accuracy of 99.44%.

Mutual Mode Entropy Based Coupling Analysis of Electroencephalogram

In this paper, coupling measure based on mutual mode entropy (MME) was applied to calculating the statistical complexity of the basic alpha rhythm extracted from the electroencephalogram (EEG) signals, which involved two groups of people, the teenager and the adult. The results show that the alpha rhythms extracted from the adult has a higher MME, which means the indication of higher statistical complexity. The following Independent Sample T Test proved that above-mentioned analysis could disclose significant differences among these two signals’ complexity.

A time-series prediction approach for feature extraction in a brain-computer interface

This paper presents a feature extraction procedure (FEP) for a brain-computer interface (BCI) application where features are extracted from the electroencephalogram (EEG) recorded from subjects performing right and left motor imagery. Two neural networks (NNs) are trained to perform one-step-ahead predictions for the EEG time-series data, where one NN is trained on right motor imagery and the other on left motor imagery. Features are derived from the power (mean squared) of the prediction error or the power of the predicted signals. All features are calculated from a window through which all predicted signals pass. Separability of features is achieved due to the morphological differences of the EEG signals and each NNs specialization to the type of data on which it is trained. Linear discriminant analysis (LDA) is used for classification. This FEP is tested on three subjects off-line and classification accuracy (CA) rates range between 88% and 98%. The approach compares favorably to a well-known adaptive autoregressive (AAR) FEP and also a linear AAR model based prediction approach.

Comparative analysis of the computational geometry and neural network classification methods for person identification purposes via the EEG : Part 1

In this paper we address the problem of person identification via features extracted from the electroencephalogram (EEG). For this problem, our work investigates appropriate feature extraction and classification methods. Our previous research into feature extraction produced a high number variety of specific EEG features, of varying dimensionalities. However, it is known that traditional classification methods such as neural networks are inefficient at classifying this type of feature. For this reason, in the present paper, we develop a heuristic classifier based on Computational Geometry algorithms. In order to exhibit the potential of the proposed method we compare it with a conventional Radial Basis Function neural classifier using the same (pre-) processing feature extraction methods in both cases.

Automated detection of a preseizure state based on a decrease in synchronization in intracranial electroencephalogram recordings from epilepsy patients.

The question whether information extracted from the electroencephalogram (EEG) of epilepsy patients can be used for the prediction of seizures has recently attracted much attention. Several studies have reported evidence for the existence of a preseizure state that can be detected using different measures derived from the theory of dynamical systems. Most of these studies, however, have neglected to sufficiently investigate the specificity of the observed effects or suffer from other methodological shortcomings. In this paper we present an automated technique for the detection of a preseizure state from EEG recordings using two different measures for synchronization between recording sites, namely, the mean phase coherence as a measure for phase synchronization and the maximum linear cross correlation as a measure for lag synchronization. Based on the observation of characteristic drops in synchronization prior to seizure onset, we used this phenomenon for the characterization of a preseizure state and its distinction from the remaining seizure-free interval. After optimizing our technique on a group of 10 patients with temporal lobe epilepsy we obtained a successful detection of a preseizure state prior to 12 out of 14 analyzed seizures for both measures at a very high specificity as tested on recordings from the seizure-free interval. After checking for in-sample overtraining via cross validation, we applied a surrogate test to validate the observed predictability. Based on our results, we discuss the differences of the two synchronization measures in terms of the dynamics underlying seizure generation in focal epilepsies.

Person Identification from the EEG using Nonlinear Signal Classification

This paper focusses on the person identification problem based on features extracted from the ElectroEncephaloGram (EEG). A bilinear rather than a purely linear model is fitted on the EEG signal, prompted by the existence of non-linear components in the EEG signal--a conjecture already investigated in previous research works. The novelty of the present work lies in the comparison between the linear and the bilinear results, obtained from real field EEG data, aiming towards identification of healthy subjects rather than classification of pathological cases for diagnosis. The EEG signal of a, in principle, healthy individual is processed via (non)linear (AR, bilinear) methods and classified by an artificial neural network classifier. Experiments performed on real field data show that utilization of the bilinear model parameters as features improves correct classification scores at the cost of increased complexity and computations. Results are seen to be statistically significant at the 99.5% level of significance, via the chi 2 test for contingency. The results obtained in the present study further corroborate existing research, which shows evidence that the EEG carries individual-specific information, and that it can be successfully exploited for purposes of person identification and authentication.

Performance differences in reaction tasks are reflected in event-related brain potentials (ERPs)

Event-related potentials (ERPs), which can be extracted from the electroencephalogram (EEG), are assumed to reflect distinct cognitive processes in real time. Hence ERP analysis could be used in cognitive ergonomics as a tool to specify, for example, bottlenecks or sources of individual performance differences. Such specific results may be helpful to change the tasks or train the subjects specifically. In the present exploratory study, the authors investigated whether subjects with large spontaneous differences in performance accuracy, as defined by their error rates in a speeded binary choice reaction task, also differ in the structure of their ERPs. The ten subjects were divided post hoc into two groups with relatively low (about 6%) and high (about 20%) error rates. While the reaction times were not significantly different for both groups, the ERPs revealed clear group differences. First, large differences were seen in the late part of the contingent negative variation (late CNV), which is assumed to reflect preparatory processes. Subjects with few errors (‘GOOD’) had a large late CNV, while subjects with many errors (‘POOR’) showed virtually no late CNV. Second, the late P300-subcomponent (which is related to response identification) was smaller and delayed for POOR compared to GOOD subjects. Finally, the ERP shows signs of poor movement control in POOR subjects. The high error rate of POOR subjects can hence be explained by: (1) their insufficient preparation for the next trial (small late CNV), which impaired response identification (small and delayed late P300 subcomponent); and (2) their poor movement control. These interpretations have to be regarded as preliminary and should be validated with larger groups of subjects. In conclusion, the main reasons for the profound performance differences between the groups, namely differential preparation and movement control, could be elucidated by ERP analysis. A potential ergonomics application of these results is that they suggest specific strategies (for example, a preparation and motor control training) to improve the performance of POOR subjects in comparable work conditions.