A Channel-Fused Dense Convolutional Network for EEG-Based Emotion Recognition

Human emotion recognition could greatly contribute to human–computer interaction with promising applications in artificial intelligence. One of the challenges in recognition tasks is learning effective representations with stable performances from electroencephalogram (EEG) signals. In this article, we propose a novel deep-learning framework, named channel-fused dense convolutional network, for EEG-based emotion recognition. First, we use a 1-D convolution layer to receive weighted combinations of contextual features along the temporal dimension from EEG signals. Next, inspired by state-of-the-art object classification techniques, we employ 1-D dense structures to capture electrode correlations along the spatial dimension. The developed algorithm is capable of handling temporal dependencies and electrode correlations with the effective feature extraction from noisy EEG signals. Finally, we perform extensive experiments based on two popular EEG emotion datasets. Results indicate that our framework achieves prominent average accuracies of 90.63% and 92.58% on the SEED and DEAP datasets, respectively, which both receive better performances than most of the compared studies. The novel model provides an interpretable solution with excellent generalization capacity for broader EEG-based classification tasks.

I. Introduction

Human emotion has a great impact on our daily activities, which is concerned with various actions, such as relaxation, work, and entertainment. There are increasing research interests in the relationship between emotions and physiological functions [1]. It has been found that positive emotions could reflect pleasurable engagement, and are beneficial for human health and attitude [2]. However, accompanied with complaints of physical symptoms, negative emotions may adversely influence mental health and even cause serious psychological problems [3]. As the information explosion through social channels, it is quite challenging to reveal one’s emotional clues. Recently, affective computing (AC) has emerged under the demand of deep knowledge and reasonable utilization for emotion [4], [5]. It is a promising area of research that has attracted increasing attentions from numerous cross-curricular fields, ranging from neuroscience to computer engineering. Emotion would be subtly influenced by multiple external and psychological factors, and is a combination of time, space, experience, and cultural background [6]. This aggravates the difficulties for emotion recognition research. Although great efforts have been made to explore the mechanisms and methods for emotion recognition [7], due to the intricate external patterns, effective emotion recognition methods are still in high demand for many technological applications.