基于3D特征融合与轻量化CNN的情绪EEG识别OA
EEG Emotion Recognition Based on 3D Feature Fusion and Lightweight CNN
情绪变化可引起头皮脑电信号的改变,基于脑电信号的情绪识别是近年来情绪研究的一个重要方向.为此,提出一种基于 3D 特征融合与轻量化卷积神经网络的情绪EEG识别方法,使用2 s窗口的3D特征图作为输入,并根据效价和唤醒提供情绪状态作为输出.在DEAP公开数据集上对所提方法进行受试者依赖实验,结果表明情绪识别性能评估效价和唤醒识别准确率分别为(97.08±0.32)%和(96.78±0.34)%.所提方法具有较高的情绪识别准确度和较低的计算复杂度,适用于实际场景中的情绪识别.
Emotional changes can cause changes in scalp EEG signals,and emotion recognition based on EEG signals has become an impor-tant direction in emotional research in recent years.To this end,a sentiment EEG recognition method based on 3D feature fusion and light-weight convolutional neural network is proposed,using a 2D window 3D feature map as input and providing emotional states as output based on valence and arousal.A subject dependent experiment was conducted on the DEAP public dataset,and the results showed that the evalua-tion validity of emotion recognition performance and the accuracy of wake-up recognition were(97.08±0.32)%and(96.78±0.34)%,respec-tively.The proposed method has high accuracy in emotion recognition and low computational complexity,making it suitable for emotion recog-nition in practical scenarios.
陈紫扬;随力;胡磊
上海理工大学 健康科学与工程学院,上海 200093
计算机与自动化
情绪识别卷积神经网络脑电信号特征融合轻量化模型
emotion recognitionconvolutional neural networksEEG signalsfeature fusionlightweight model
《软件导刊》 2024 (006)
38-43 / 6
上海理工大学科技发展项目(2019KJFZ239,2020KJFZ232)
评论