EEG-based emotion recognition using deep learning network with principal component based covariate shift adaptation.
ABSTRACT: Automatic emotion recognition is one of the most challenging tasks. To detect emotion from nonstationary EEG signals, a sophisticated learning algorithm that can represent high-level abstraction is required. This study proposes the utilization of a deep learning network (DLN) to discover unknown feature correlation between input signals that is crucial for the learning task. The DLN is implemented with a stacked autoencoder (SAE) using hierarchical feature learning approach. Input features of the network are power spectral densities of 32-channel EEG signals from 32 subjects. To alleviate overfitting problem, principal component analysis (PCA) is applied to extract the most important components of initial input features. Furthermore, covariate shift adaptation of the principal components is implemented to minimize the nonstationary effect of EEG signals. Experimental results show that the DLN is capable of classifying three different levels of valence and arousal with accuracy of 49.52% and 46.03%, respectively. Principal component based covariate shift adaptation enhances the respective classification accuracy by 5.55% and 6.53%. Moreover, DLN provides better performance compared to SVM and naive Bayes classifiers.
Project description:The electroencephalogram (EEG) has great attraction in emotion recognition studies due to its resistance to deceptive actions of humans. This is one of the most significant advantages of brain signals in comparison to visual or speech signals in the emotion recognition context. A major challenge in EEG-based emotion recognition is that EEG recordings exhibit varying distributions for different people as well as for the same person at different time instances. This nonstationary nature of EEG limits the accuracy of it when subject independency is the priority. The aim of this study is to increase the subject-independent recognition accuracy by exploiting pretrained state-of-the-art Convolutional Neural Network (CNN) architectures. Unlike similar studies that extract spectral band power features from the EEG readings, raw EEG data is used in our study after applying windowing, pre-adjustments and normalization. Removing manual feature extraction from the training system overcomes the risk of eliminating hidden features in the raw data and helps leverage the deep neural network's power in uncovering unknown features. To improve the classification accuracy further, a median filter is used to eliminate the false detections along a prediction interval of emotions. This method yields a mean cross-subject accuracy of 86.56% and 78.34% on the Shanghai Jiao Tong University Emotion EEG Dataset (SEED) for two and three emotion classes, respectively. It also yields a mean cross-subject accuracy of 72.81% on the Database for Emotion Analysis using Physiological Signals (DEAP) and 81.8% on the Loughborough University Multimodal Emotion Dataset (LUMED) for two emotion classes. Furthermore, the recognition model that has been trained using the SEED dataset was tested with the DEAP dataset, which yields a mean prediction accuracy of 58.1% across all subjects and emotion classes. Results show that in terms of classification accuracy, the proposed approach is superior to, or on par with, the reference subject-independent EEG emotion recognition studies identified in literature and has limited complexity due to the elimination of the need for feature extraction.
Project description:Constructing a robust emotion-aware analytical framework using non-invasively recorded electroencephalogram (EEG) signals has gained intensive attentions nowadays. However, as deploying a laboratory-oriented proof-of-concept study toward real-world applications, researchers are now facing an ecological challenge that the EEG patterns recorded in real life substantially change across days (i.e., day-to-day variability), arguably making the pre-defined predictive model vulnerable to the given EEG signals of a separate day. The present work addressed how to mitigate the inter-day EEG variability of emotional responses with an attempt to facilitate cross-day emotion classification, which was less concerned in the literature. This study proposed a robust principal component analysis (RPCA)-based signal filtering strategy and validated its neurophysiological validity and machine-learning practicability on a binary emotion classification task (happiness vs. sadness) using a five-day EEG dataset of 12 subjects when participated in a music-listening task. The empirical results showed that the RPCA-decomposed sparse signals (RPCA-S) enabled filtering off the background EEG activity that contributed more to the inter-day variability, and predominately captured the EEG oscillations of emotional responses that behaved relatively consistent along days. Through applying a realistic add-day-in classification validation scheme, the RPCA-S progressively exploited more informative features (from 12.67 ± 5.99 to 20.83 ± 7.18) and improved the cross-day binary emotion-classification accuracy (from 58.31 ± 12.33% to 64.03 ± 8.40%) as trained the EEG signals from one to four recording days and tested against one unseen subsequent day. The original EEG features (prior to RPCA processing) neither achieved the cross-day classification (the accuracy was around chance level) nor replicated the encouraging improvement due to the inter-day EEG variability. This result demonstrated the effectiveness of the proposed method and may shed some light on developing a realistic emotion-classification analytical framework alleviating day-to-day variability.
Project description:Human emotion recognition has been a major field of research in the last decades owing to its noteworthy academic and industrial applications. However, most of the state-of-the-art methods identified emotions after analyzing facial images. Emotion recognition using electroencephalogram (EEG) signals has got less attention. However, the advantage of using EEG signals is that it can capture real emotion. However, very few EEG signals databases are publicly available for affective computing. In this work, we present a database consisting of EEG signals of 44 volunteers. Twenty-three out of forty-four are females. A 32 channels CLARITY EEG traveler sensor is used to record four emotional states namely, happy, fear, sad, and neutral of subjects by showing 12 videos. So, 3 video files are devoted to each emotion. Participants are mapped with the emotion that they had felt after watching each video. The recorded EEG signals are considered further to classify four types of emotions based on discrete wavelet transform and extreme learning machine (ELM) for reporting the initial benchmark classification performance. The ELM algorithm is used for channel selection followed by subband selection. The proposed method performs the best when features are captured from the gamma subband of the FP1-F7 channel with 94.72% accuracy. The presented database would be available to the researchers for affective recognition applications.
Project description:Emotions are fundamental for human beings and play an important role in human cognition. Emotion is commonly associated with logical decision making, perception, human interaction, and to a certain extent, human intelligence itself. With the growing interest of the research community towards establishing some meaningful “emotional” interactions between humans and computers, the need for reliable and deployable solutions for the identification of human emotional states is required. Recent developments in using electroencephalography (EEG) for emotion recognition have garnered strong interest from the research community as the latest developments in consumer-grade wearable EEG solutions can provide a cheap, portable, and simple solution for identifying emotions. Since the last comprehensive review was conducted back from the years 2009 to 2016, this paper will update on the current progress of emotion recognition using EEG signals from 2016 to 2019. The focus on this state-of-the-art review focuses on the elements of emotion stimuli type and presentation approach, study size, EEG hardware, machine learning classifiers, and classification approach. From this state-of-the-art review, we suggest several future research opportunities including proposing a different approach in presenting the stimuli in the form of virtual reality (VR). To this end, an additional section devoted specifically to reviewing only VR studies within this research domain is presented as the motivation for this proposed new approach using VR as the stimuli presentation device. This review paper is intended to be useful for the research community working on emotion recognition using EEG signals as well as for those who are venturing into this field of research.
Project description:The hierarchical interaction between electrical signals of the brain and heart is not fully understood. We hypothesized that the complexity of cardiac electrical activity can be used to predict changes in encephalic electricity after stress. Most methods for analyzing the interaction between the heart rate variability (HRV) and electroencephalography (EEG) require a computation-intensive mathematical model. To overcome these limitations and increase the predictive accuracy of human relaxing states, we developed a method to test our hypothesis. In addition to routine linear analysis, multiscale entropy and detrended fluctuation analysis of the HRV were used to quantify nonstationary and nonlinear dynamic changes in the heart rate time series. Short-time Fourier transform was applied to quantify the power of EEG. The clinical, HRV, and EEG parameters of postcatheterization EEG alpha waves were analyzed using change-score analysis and generalized additive models. In conclusion, the complexity of cardiac electrical signals can be used to predict EEG changes after stress.
Project description:Color is a perceptual stimulus that has a significant impact on improving human emotion and memory. Studies have revealed that colored multimedia learning materials (MLMs) have a positive effect on learner's emotion and learning where it was assessed by subjective/objective measurements. This study aimed to quantitatively assess the influence of colored MLMs on emotion, cognitive processes during learning, and long-term memory (LTM) retention using electroencephalography (EEG). The dataset consisted of 45 healthy participants, and MLMs were designed in colored or achromatic illustrations to elicit emotion and that to assess its impact on LTM retention after 30-min and 1-month delay. The EEG signal analysis was first started to estimate the effective connectivity network (ECN) using the phase slope index and expand it to characterize the ECN pattern using graph theoretical analysis. EEG results showed that colored MLMs had influences on theta and alpha networks, including (1) an increased frontal-parietal connectivity (top-down processing), (2) a larger number of brain hubs, (3) a lower clustering coefficient, and (4) a higher local efficiency, indicating that color influences information processing in the brain, as reflected by ECN, together with a significant improvement in learner's emotion and memory performance. This is evidenced by a more positive emotional valence and higher recall accuracy for groups who learned with colored MLMs than that of achromatic MLMs. In conclusion, this paper demonstrated how the EEG ECN parameters could help quantify the influences of colored MLMs on emotion and cognitive processes during learning.
Project description:Estimating the depth of anaesthesia (DoA) in operations has always been a challenging issue due to the underlying complexity of the brain mechanisms. Electroencephalogram (EEG) signals are undoubtedly the most widely used signals for measuring DoA. In this paper, a novel EEG-based index is proposed to evaluate DoA for 24 patients receiving general anaesthesia with different levels of unconsciousness. Sample Entropy (SampEn) algorithm was utilised in order to acquire the chaotic features of the signals. After calculating the SampEn from the EEG signals, Random Forest was utilised for developing learning regression models with Bispectral index (BIS) as the target. Correlation coefficient, mean absolute error, and area under the curve (AUC) were used to verify the perioperative performance of the proposed method. Validation comparisons with typical nonstationary signal analysis methods (i.e., recurrence analysis and permutation entropy) and regression methods (i.e., neural network and support vector machine) were conducted. To further verify the accuracy and validity of the proposed methodology, the data is divided into four unconsciousness-level groups on the basis of BIS levels. Subsequently, analysis of variance (ANOVA) was applied to the corresponding index (i.e., regression output). Results indicate that the correlation coefficient improved to 0.72 ± 0.09 after filtering and to 0.90 ± 0.05 after regression from the initial values of 0.51 ± 0.17. Similarly, the final mean absolute error dramatically declined to 5.22 ± 2.12. In addition, the ultimate AUC increased to 0.98 ± 0.02, and the ANOVA analysis indicates that each of the four groups of different anaesthetic levels demonstrated significant difference from the nearest levels. Furthermore, the Random Forest output was extensively linear in relation to BIS, thus with better DoA prediction accuracy. In conclusion, the proposed method provides a concrete basis for monitoring patients' anaesthetic level during surgeries.
Project description:Emotion regulation deficits are commonly observed in social anxiety disorder (SAD). We used manifold-learning to learn the phase-space connectome manifold of EEG brain dynamics in twenty SAD participants and twenty healthy controls. The purpose of the present study was to utilize manifold-learning to understand EEG brain dynamics associated with emotion regulation processes. Our emotion regulation task (ERT) contains three conditions: Neutral, Maintain and Reappraise. For all conditions and subjects, EEG connectivity data was converted into series of temporally-consecutive connectomes and aggregated to yield this phase-space manifold. As manifold geodesic distances encode intrinsic geometry, we visualized this space using its geodesic-informed minimum spanning tree and compared neurophysiological dynamics across conditions and groups using the corresponding trajectory length. Results showed that SAD participants had significantly longer trajectory lengths during Neutral and Maintain. Further, trajectory lengths during Reappraise were significantly associated with the habitual use of reappraisal strategies, while Maintain trajectory lengths were significantly associated with the negative affective state during Maintain. In sum, an unsupervised connectome manifold-learning approach can reveal emotion regulation associated phase-space features of brain dynamics.
Project description:Emotion classification based on brain-computer interface (BCI) systems is an appealing research topic. Recently, deep learning has been employed for the emotion classifications of BCI systems and compared to traditional classification methods improved results have been obtained. In this paper, a novel deep neural network is proposed for emotion classification using EEG systems, which combines the Convolutional Neural Network (CNN), Sparse Autoencoder (SAE), and Deep Neural Network (DNN) together. In the proposed network, the features extracted by the CNN are first sent to SAE for encoding and decoding. Then the data with reduced redundancy are used as the input features of a DNN for classification task. The public datasets of DEAP and SEED are used for testing. Experimental results show that the proposed network is more effective than conventional CNN methods on the emotion recognitions. For the DEAP dataset, the highest recognition accuracies of 89.49% and 92.86% are achieved for valence and arousal, respectively. For the SEED dataset, however, the best recognition accuracy reaches 96.77%. By combining the CNN, SAE, and DNN and training them separately, the proposed network is shown as an efficient method with a faster convergence than the conventional CNN.
Project description:Motivation. Anomaly EEG detection is a long-standing problem in analysis of EEG signals. The basic premise of this problem is consideration of the similarity between two nonstationary EEG recordings. A well-established scheme is based on sequence matching, typically including three steps: feature extraction, similarity measure, and decision-making. Current approaches mainly focus on EEG feature extraction and decision-making, and few of them involve the similarity measure/quantification. Generally, to design an appropriate similarity metric, that is compatible with the considered problem/data, is also an important issue in the design of such detection systems. It is however impossible to directly apply those existing metrics to anomaly EEG detection without any consideration of domain specificity. Methodology. The main objective of this work is to investigate the impacts of different similarity metrics on anomaly EEG detection. A few metrics that are potentially available for the EEG analysis have been collected from other areas by a careful review of related works. The so-called power spectrum is extracted as features of EEG signals, and a null hypothesis testing is employed to make the final decision. Two indicators have been used to evaluate the detection performance. One is to reflect the level of measured similarity between two compared EEG signals, and the other is to quantify the detection accuracy. Results. Experiments were conducted on two data sets, respectively. The results demonstrate the positive impacts of different similarity metrics on anomaly EEG detection. The Hellinger distance (HD) and Bhattacharyya distance (BD) metrics show excellent performances: an accuracy of 0.9167 for our data set and an accuracy of 0.9667 for the Bern-Barcelona EEG data set. Both of HD and BD metrics are constructed based on the Bhattacharyya coefficient, implying the priority of the Bhattacharyya coefficient when dealing with the highly noisy EEG signals. In future work, we will exploit an integrated metric that combines HD and BD for the similarity measure of EEG signals.