Self supervised learning based emotion recognition using physiological signals.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Emotion recognition is crucial for advanced human-machine interaction.
• Physiological signals, particularly Electroencephalogram (EEG), offer objective insights into human emotional states.
• Current labeled EEG datasets for emotion recognition are often limited in size, hindering model development.

Purpose of the Study:

• To investigate the efficacy of self-supervised learning for emotion recognition using EEG signals.
• To address the challenge of limited labeled data in EEG-based emotion recognition.

Main Methods:

• Utilized self-supervised learning techniques on EEG data.
• Employed three pre-defined tasks to generate pseudo-labels and extract features from the data's inherent structure.
• Focused on learning feature representations without relying on manual annotations.

Main Results:

• Self-supervised learning successfully learned effective feature representations from unlabeled EEG data.
• The extracted features are applicable to downstream emotion recognition tasks.
• Demonstrated the potential of unsupervised methods in this domain.

Conclusions:

• Self-supervised learning is a viable and powerful approach for EEG-based emotion recognition.
• This method enables the utilization of large unlabeled EEG datasets, overcoming the bottleneck of data scarcity.
• The findings pave the way for more robust and scalable emotion recognition systems in human-machine interaction.