Luancheng, Shi Yun

(College of Biomedical Engineering, Chengdu University)

Abstract：

With the continuous advancement and widespread application of human-computer interaction technology, accurate recognition of user emotions has become increasingly important. Emotion recognition technology has shown great potential in multiple fields, including medical diagnosis, traffic safety, and education. And emotion recognition based on EEG has become a popular research direction in the field of emotion recognition. Firstly, this article introduces the basic concepts of emotion continuity and emotion dispersion models, summarizes commonly used publicly available EEG datasets, and compares and analyzes the size of the datasets, emotion labels, and their impact on EEG emotion recognition tasks. Secondly, as the impact of different frequency bands and EEG channels on emotion recognition varies, the research team has compiled relevant studies on key frequency bands and channels for EEG emotion recognition, and summarized the range of key frequency bands for EEG emotion recognition and the location of EEG channels with rich emotional information from literature. Next, four types of EEG emotion features are introduced and corresponding feature extraction methods are provided. The difficulty of extracting various EEG features and their current application effects in emotion recognition are also pointed out. Then, the data augmentation techniques and attention mechanisms in deep learning based EEG emotion recognition were elaborated, pointing out the mainstream trends of data augmentation techniques and the types of artificial emotion features generated. A comparative analysis was conducted on the ways and focuses of various attention mechanisms. Data augmentation technology is used to solve the problem of insufficient EEG data, and attention mechanism plays a key role in improving the accuracy of emotion recognition. Finally, a certain outlook was made on the universality of future EEG emotion recognition models and the research direction of EEG acquisition devices.

Key Words：

emotion recognition; EEG signals; deep learning; data augmentation; attention mechanism