A Review on EEG-based Multimodal Learning for Emotion Recognition

Abstract

Emotion recognition from electroencephalography (EEG) signal is crucial for human-computer interaction, yet poses significant challenges. While various techniques exist for detecting emotions through EEG signals, contemporary studies have explored multimodal approaches as a promising advancement. This paper offers an overview of multimodal techniques in EEG-based emotion identification and discusses recent literature in this area. But these models are computational hungry, which is necessary to address through our research, highlighting the need for further research. A relatively unexplored avenue is combining EEG data with behavioral modalities, considering unpredictable levels of reliability. The suggested review examines the strengths and pitfalls of existing multimodal emotion recognition approaches from 2017 to 2024. Key contributions include a systematic survey on EEG features, exploration of EEG integration with behavioral modalities, and investigation of fusion methods like conventional and deep learning techniques. Finally, key challenges and future research directions in implementing multi-modal emotion identification systems.