Distributed Driver Emotion Prediction for Intelligent 6G Internet of Vehicles

Accurate driver emotion prediction is critical for enhancing safety and efficiency in 6G-enabled Intelligent Transportation Systems (ITS). In real-world ITS, drivers may face complex situations, and as a results they may express complex emotion. However, traditional driver emotion analysis struggles to capture the complexity of such blended emotions. To address this challenge, we apply emotion distributions to represent complex driver emotions. Besides, we apply the Label Enhancement (LE) to transform logical emotion labels into more informative emotion distributions, which reflect the varying relevance of multiple emotions. We propose in this paper a novel distributed LE method, called Distributed Label Enhancement with Label Manifold (DLEM). It leverages the manifold structure of logical emotions to enhance them into emotion distributions. DLEM is well-suited for edge and large-scale Internet-of-Vehicles (IoV) systems. Additionally, we design the Distributed Emotion Distribution Learning (DEDL) to learn and predict driver emotions from such enhanced emotion distributions. Finally, we conduct experiments on three large-scale emotion datasets. The experimental results show that DLEM significantly outperforms several state-of-the-art approaches and achieves the best performance.