| 杨超宇,余维哲,卢绍田,孙成圆,武柏祥.基于CNN的脑电信号情绪识别模型研究[J].井冈山大学自然版,2024,45(1):76-83 |
| 基于CNN的脑电信号情绪识别模型研究 |
| EMOTION RECOGNITION MODEL OF EEG SIGNALS BASED ON CNN |
| 投稿时间：2023-10-09 修订日期：2023-12-11 |
| DOI：10.3969/j.issn.1674-8085.2024.01.001 |
| 中文关键词: 脑电波 情绪识别 CNN 脑电信号 |
| 英文关键词: brain wave emotional recognition CNN EEG signal |
| 基金项目:国家自然科学基金项目(61873004)；安徽省大学生创新创业训练计划项目(S202210361269) |
| 作者 | 单位 | | 杨超宇 | 安徽理工大学人工智能学院, 安徽, 淮南 232001 | | 余维哲 | 安徽理工大学人工智能学院, 安徽, 淮南 232001 | | 卢绍田 | 安徽理工大学人工智能学院, 安徽, 淮南 232001 | | 孙成圆 | 安徽理工大学人工智能学院, 安徽, 淮南 232001 | | 武柏祥 | 安徽理工大学人工智能学院, 安徽, 淮南 232001 |
|
| 摘要点击次数: 147 |
| 全文下载次数: 503 |
| 中文摘要: |
| 针对现有深度学习模型在情绪识别方面种类少且准确率低的问题，采集并建立了脑电波信号数据集，提出了一种基于CNN的脑电波的智能多情绪识别模型，利用多层卷积神经网络提取脑电信号情感特征，在批归一化层和激活函数中引入非线性特性，构建了两层全连接神经网络，实现了情绪特征中积极、中性和悲伤的分类。实验结果表明，提出的模型复杂度低且分类准确率达到了81.43%，明显高于SVM、LSTM、VGGNet模型，证明了该模型的简洁性和高效性。 |
| 英文摘要: |
| In response to the limited variety and low accuracy of existing deep learning models for emotion recognition, a dataset of electroencephalogram (EEG) signals was collected and established, and an intelligent multi-emotion recognition model based on Convolutional Neural Networks (CNNs) was developed. The model utilizes multiple layers of convolutional neural networks to extract emotional features from EEG signals. Non-linear characteristics are introduced through batch normalization layers and activation functions. Additionally, a two-layer fully connected neural network is designed to classify emotional features into positive, neutral, and sad categories. The experimental results demonstrate that the proposed model exhibits low complexity and achieves a classification accuracy of 81.43%, surpassing SVM, LSTM, and VGGNet models. This confirms the efficiency and simplicity of the proposed model. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
