Recognition of Emotional States using Multiscale Information Analysis of High Frequency EEG Oscillations

Abstract

Exploring the manifestation of emotion in electroencephalogram (EEG) signals is helpful for improving the accuracy of emotion recognition. This paper introduced the novel features based on the multiscale information analysis (MIA) of EEG signals for distinguishing emotional states in four dimensions based on Russell's circumplex model. The algorithms were applied to extract features on the DEAP database, which included multiscale EEG complexity index in the time domain, and ensemble empirical mode decomposition enhanced energy and fuzzy entropy in the frequency domain. The support vector machine and cross validation method were applied to assess classification accuracy. The classification performance of MIA methods (accuracy = 62.01%, precision = 62.03%, recall/sensitivity = 60.51%, and specificity = 82.80%) was much higher than classical methods (accuracy = 43.98%, precision = 43.81%, recall/sensitivity = 41.86%, and specificity = 70.50%), which extracted features contain similar energy based on a discrete wavelet transform, fractal dimension, and sample entropy. In this study, we found that emotion recognition is more associated with high frequency oscillations (51-100Hz) of EEG signals rather than low frequency oscillations (0.3-49Hz), and the significance of the frontal and temporal regions are higher than other regions. Such information has predictive power and may provide more insights into analyzing the multiscale information of high frequency oscillations in EEG signals. Entropy 2019, 21, 609 2 of 21 are extracted as features for emotion recognition by Lin et al. [3]. Petrantonakis et al. [4] introduced higher-order crossing features to capture the oscillatory pattern of EEG. The Hjorth parameters are developed and used to distinguish emotions [5]. Liu et al. [6] proposed the fractal dimension (FD) based algorithm on quantification of basic emotions and described its implementation as feedback in 3D virtual environments. Several entropy-based metrics of signal complexity have already been proposed for discriminating emotional states. Hosseini et al. [7] applied two entropy metrics (approximate and wavelet entropy) to discriminate between two emotional states (calm-neutral and negative-excited) in response to viewing sequences of emotion-inducing pictures and achieved 73.25% classification accuracy. Jie et al. [8] applied sample entropy (SE) to EEG data obtained from two binary emotion recognition tasks (positive vs. negative emotion both with high arousal, and music clips with different arousal levels) and achieved 80.43% and 79.11% classification performance. Murugappan et al. [9] used the discrete wavelet transform (DWT) to divide the EEG signal into several bands. Then they calculated features based on these bands. Despite the fact that some encouraging progress has been made, developing the best combination of feature extraction and classification methods still require further research.A promising development on EEG signals for emotion recognition is multiscale analysis, including correlation dimensi...

Human Emotion Recognition: Review of Sensors and Methods

“…The EEG input signals were provided by the DEAP dataset [61] From information provided in Table 2 it is seen that main part of researches are focused on the development of more advanced methods for emotion recognition from EEG signals. For this purpose, it is generally not required to provide a real experiment in order to validate the proposed method, because free databases are available with recorded EEG signals under known conditions.…”

Frontal EEG-Based Multi-Level Attention States Recognition Using Dynamical Complexity and Extreme Gradient Boosting

“…Rezaeezadeh et al (2020) focused on entropy measures, including univariate features from individual EEG channels and multivariate features from brain lobes, to diagnosis Attention Deficit Hyperactivity Disorder. Recent studies have shown that applying nonlinear multiscale information analysis to EEG can provide new information about the complex dynamics of brain cognitive function, such as emotion recognition (Gao et al, 2019). Ke et al (2014) conducted two experiments instructing all subjects to perform tasks with three different levels of attention (i.e., attention, no attention, and rest).…”

Conditional Adversarial Domain Adaptation Neural Network for Motor Imagery EEG Decoding

“…Firstly, a batch of EEG signal matrixes consisting of and are transformed into the high-level discriminative features and by the feature extractor. Then, these features are typically fed into the classifier to obtain the conditional probabilities vectors and with the softmax function [ 7 ]. Finally, the conditional features are calculated on the basis of the features and probabilities generated above, and the conditional discriminator is introduced to distinguish the EEG data from the source domain or target domain, in which the domain labels of the target dataset are set to 0 and, correspondingly, the domain label of the source is set to 1.…”

“…These methods mainly include two stages. Firstly, discriminative features, such as entropy feature sets [ 6 ], the frequency band power [ 7 ], and the filter band common spatial pattern (FBCSP) [ 8 , 9 ] are extracted from each EEG trial. Then, these informative features are fed into classifiers, including a support vector machine (SVM) and random forest, to generate the final recognition result [ 10 , 11 ].…”