基于卷积神经网络的脑电信号情绪识别方法研究
首发时间:2020-06-05
摘要:情绪是反映人的心理状态的指标,情绪的识别对于人们的研究和生活都已经有了很重要的意义。脑电信号常被用于情绪识别领域,常使用频空特征进行识别。考虑到不同种类特征之间具有较强的互补性,本文提出了将频、空两种特征相结合的方法。同时,本文提出的伪3D卷积的计算模式,降低了模型的计算成本和内存需求。模型引入的残差结构,解决了网络加深带来的梯度消失和梯度爆炸的问题。随后,本文在SEED数据集上进行了脑电信号的情绪识别实验,实验结果表明该方法达到了出色性能。
For information in English, please click here
Research on EEG Emotion Recognition Method Based on Convolutional Neural Network
Abstract:Emotion is an indicator of evaluating a person\'s psychological state, and emotion recognition has significant meanings for people\'s researches and life. Electroencephalogram (EEG) is often used in the field of emotion recognition.Frequency and spatial features of EEG signals are often used for emotion recognition. Considering the strong complementarity between different kinds of features, this paper proposes a method using combination of frequency features and spatial features. At the same time, the computational method of pseudo 3D convolution proposed in this paper reduces the computational cost and memory requirements of model. The residual structure introduced by the model solves the problem of gradient vanishing and expolding caused by network deepening. Subsequently, the emotion recognition experiment of EEG signals is carried out on the SEED dataset in this paper. The experimental results show that the proposed method achieves excellent performance.
Keywords: deep learning emotion recogniton physiological signals processing
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于卷积神经网络的脑电信号情绪识别方法研究
评论
全部评论0/1000