EEG Emotion Recognition via Temporal Asynchronous Alignment Contrastive Learning
A new framework called Temporal Asynchronous Alignment-based Contrastive Learning (TA2CL) has been proposed for cross-subject EEG emotion recognition. The method addresses the issue of temporal misalignment in neural responses across different subjects by adapting the late interaction mechanism from ColBERT, a natural language processing model. Instead of traditional global hard alignment, TA2CL uses fine-grained local matching to improve similarity calculations. The research was published on arXiv with ID 2605.22379v1.
Key facts
- TA2CL is a new framework for cross-subject EEG emotion recognition.
- It draws inspiration from ColBERT's late interaction mechanism in NLP.
- The method transforms global hard alignment into fine-grained local matching.
- It addresses temporal misalignment of responses among different subjects.
- The paper is available on arXiv with ID 2605.22379v1.
- The research focuses on improving similarity calculation strategies.
- EEG-based emotion recognition is valued for its objectivity and high temporal resolution.
- Existing methods often overlook temporal misalignment in cross-subject scenarios.
Entities
Institutions
- arXiv