Adaptive Emotion Detection Using the Valence-Arousal-Dominance Model and EEG Brain Rhythmic Activity Changes in Relevant Brain Lobes

Abstract

Contrary to popular belief that only the frontal lobe is concerned with emotions, recent neuroscience experiments show that in each emotional state, some of the brain lobes fired clearly whereas some did not. Unlike most of the previous works, which focused on choosing a fixed set of electrodes to detect emotions, this work presents a novel approach that consists of investigating the cerebral activity while experiencing emotions to identify the brain lobes that are showing significant and relevant changes. The identification process of relevant brain lobes is adaptive as the brain activity varies from one person to another and differs from one lobe to another during the same emotional state. The electrodes of the selected brain lobes will be tuned before being used as resources to extract the features that are required for the proposed three phases classification process. Using a clustering technique, the tuning process excludes every electrode that hardly separates between relevant and irrelevant rhythmic brain changes. The results show that the proposed method outperforms all previous approaches. Compared to the best performance obtained by previous studies, the proposed method enhanced the accuracy of both valence and arousal dimensions by 5%. Besides, the accuracy of the dominance dimension was improved by 2%.

Channels and Features Identification: A Review and a Machine-Learning Based Model With Large Scale Feature Extraction for Emotions and ASD Classification

“…Therefore, not many data sets are considered a benchmark for the EEG-based emotions classification (Koelstra et al, 2011;Zheng and Lu, 2015). The previous works have performed a very notable effort toward emotions classification and have achieved very good results (Gannouni et al, 2020;Alturki et al, 2021). They can be broadly categorized into software or hardwarebased solutions.…”

“…Different emotions measured using the 3-D scale (valence, arousal, and dominance) and 2-D scale (valence, arousal) are depicted in Figure 3 (Gannouni et al, 2020). The happy and joy emotions are differentiated by different dominance values.…”

“…Figure 3A shows the neutral, happy, joy, sad, depressed, angry, and relaxed emotions mapped using the valence, arousal, and dominance scale. It becomes hard to label a large set of emotions on the 3-D scale (valence, arousal, and dominance) (Gannouni et al, 2020). Therefore, Russel's 2-D scale (valence, arousal) is most popularly used to quantify emotions (Russell, 1980).…”

The Influence of Mutual Assistance of Construction Workers with Different Personality Traits on Team Safety

“…Chi et al [34] pointed out that workers' unsafe behaviors are misjudgments or wrong decisions made in the cognitive process. From a physiological point of view, construction workers' cognitive status can also be assessed by valence, arousal, and dominance index in the valence-arousal-dominance (VAD) model [35], which can be measured by electroencephalography (EEG) [36]. By analyzing the three broadly influential cognitive models: Rasmussen's step-ladder model [37], Wickens et al's model of human information processing [38], and the IDAC model [39], Fang et al [31] summarized the cognitive process of construction workers' unsafe behaviors into five stages: discovering information, understanding information, thinking and responding, choosing a response, and implementing the response.…”

Research on the Emotions Based on Brain-Computer Technology: A Bibliometric Analysis and Research Agenda

“…This instrument evaluates EEG pressure features by analyzing the frontal asymmetry and uses machine learning classifiers such as SVM to classify EEG, with high classification accuracy. Gannouni et al (2020) proposed VAD model to extract, select and classify EEG signals from three dimensions of valence, arousal, and dominance, respectively.…”