Application of Multimodal Emotion Recognition in Intelligent Vehicles

Abstract

Road traffic safety is a critical global issue, as approximately 1.2 million people die from road traffic accidents annually. Driver behavior is influenced by negative emotions like anger, anxiety, and fatigue, which is a key contributing factor of road accidents. Multimodal emotion recognition, which integrates diverse data types such as images, audio, physiological signals, and vehicle data, significantly enhances the accuracy and robustness of emotion recognition systems, playing a important role in preventing potential accidents. This paper systematically analyzes representative research on multimodal emotion recognition used in automobile scenarios from 2020 to 2025, focusing on core technologies including data preprocessing, feature extraction, and fusion strategies. It critically analyzes algorithm frameworks while examining typical cases, and discusses limitations in privacy and public datasets offering potential ways of addressing them. It aims to provide comprehensive references for advancing driver emotion recognition research in intelligent vehicles. Ultimately facilitating accident prevention and optimizing vehicle interaction experiences.