基于KDCCA的脑电信号辅助外周信号的情感识别研究

Research on Emotional Recognition of Peripheral Signals Assisted by EEG Signals Based on KDCCA

多信号融合是生理信号情感识别中的重点,其中,生理信号和脑电信号(EEG)被广泛使用。但EEG信号获取困难、成本高,为更有效地使用EEG信号,提出了一种基于核化的判别型典型相关分析(KDCCA)的脑电信号辅助生理信号的情感分类方法。训练时先提取各种信号,在EEG信号的辅助下使用KDCCA创建新的判别空间,然后采用多种机器学习方法构建情感模型,最后在测试时只使用EEG信号。经实验验证,所提方法取得了更好的分类效果。

Multi signal fusion is a key focus in emotional recognition of physiological signals,among which physiological signals and electroencephalogram(EEG)signals are widely used.However,obtaining EEG signals is difficult and costly.In order to use EEG signals more effectively,a sentiment classification method based on kernel discriminant canonical correlation analysis(KDCCA)for EEG signal assisted physiological signals is proposed.During training,various signals are first extracted,and KDCCA is used to create a new discriminative space with the assistance of EEG signals.Then,various machine learning methods are used to construct sentiment models,and finally,only EEG signals are used during testing.After verification,the proposed method achieved better classification performance.