Virtual reality-enabled high-performance emotion estimation with the most significant channel pairs

Area of Science:

• Neuroscience and Human-Computer Interaction (HCI)
• Cognitive Science and Affective Computing

Background:

• Electroencephalogram (EEG) signals are crucial for assessing user experience (UX) in virtual environments, enabling real-time adaptation of metaverse systems.
• Analyzing EEG signals can optimize UX, enhance immersion, and personalize interactions, but challenges like processing costs, classification accuracy, and cybersickness in Virtual Reality (VR) persist.
• Reducing noisy and redundant information from unrelated EEG channels through effective channel selection is vital for improving HCI and UX applications.

Purpose of the Study:

• To propose a new EEG channel selection method based on phase-locking value (PLV) analysis for emotion estimation.
• To investigate interactions between EEG channels using PLV in repeated experimental tasks.
• To enhance classification performance for emotion prediction using a reduced set of EEG channels.

Main Methods:

• A novel EEG channel selection technique utilizing phase-locking value (PLV) analysis was developed.
• Frequency-based features were extracted from the selected EEG channels.
• A Multiple-Instance Learning (MIL) variant was employed for classification, dividing features into bags.
• The Random Forests (RF) algorithm was used for binary classification tasks.

Main Results:

• The proposed method demonstrated higher classification performance for emotion prediction with fewer EEG channels.
• Binary classification using the Random Forests (RF) algorithm achieved a promising accuracy rate of 99%.
• On the VREMO dataset, the method reached 99.38% accuracy for valence with all channels and 98.13% with selected channels.
• On the DEAP dataset, accuracies were 98.16% with all channels and 98.13% with selected channels.

Conclusions:

• The developed EEG channel selection method based on PLV analysis effectively reduces noisy and redundant information, leading to improved emotion prediction accuracy.
• This approach offers a promising solution for optimizing HCI and UX in immersive environments by enabling efficient and accurate analysis of EEG signals.
• The study highlights the potential of PLV-based channel selection for real-time emotion recognition in VR and metaverse applications, addressing key challenges in signal processing and classification.