Venue: IEEE Transactions on Affective Computing

Authors: Ziyu Jia, Heng Liang, Haichao Wang, Yucheng Liu, Tianzi Jiang

We propose a Cross-Modal Emotion Knowledge Distillation (CMEKD) framework that extracts the heterogeneity and interactivity between GSR and EEG signals and conveys a comprehensive fusion of multimodal features into an unimodal GSR model through knowledge distillation, thereby enhancing unimodal model performance. We employ cosine similarity-based knowledge representation and an adaptive feedback mechanism to dynamically adjust the distillation process.

Tags: Knowledge Distillation, Emotion Recognition, Cross-Modal