基于脑电图的自注意力卷积神经网络酒瘾检测模型  

A Convolutional Neural Networks Model for Alcohol Addiction Detection Based on EEG and Self-attention Mechanisms

作　　者：张东晓[1] 王佳毅 叶贵 马素凡 ZHANG Dongxiao;WANG Jiayi;YE Gui;MA Sufan(College of Science,Jimei University,Xiamen 361021,China)

机构地区：[1]集美大学理学院,福建厦门361021

出　　处：《内蒙古民族大学学报(自然科学版)》2025年第2期59-68,共10页Journal of Inner Mongolia Minzu University:Natural Sciences Edition

基　　金：国家自然科学基金项目(12271211);福建省科技创新智库课题研究项目(FJKX-2023XKB007)。

摘　　要：目前酒精成瘾主要由医生根据经验来诊断,主观性较强,使用脑电图(EEG)检测酒瘾,可以辅助医生做出客观的判断。提出一种基于注意力机制的酒瘾检测模型AADNet。AADNet由卷积模块、自注意力模块、特征增强模块和分类模块构成。卷积模块通过空间卷积和时间卷积,提取EEG信号的局部特征;自注意力模块通过空间自注意力机制和特征自注意力机制提取EEG信号的全局特征;特征增强模块进一步融合局部特征和全局特征,提取与类别强相关的特征;分类模块负责预测酒瘾的概率。实验结果表明,本文模型可以有效检测酒瘾,在公开数据集上的准确率可以达到100.00%,优于目前的大多数算法。Alcohol addiction is currently mainly diagnosed by doctors based on experience,which is highly subjective.The use of electroencephalography(EEG)to detect alcohol addiction can assist doctors in making objective judgments.This study proposes an alcohol addiction detection CNN model AADNet based on attention mechanism.AADNet consists of a convolution module,a self-attention module,a feature enhancement module and a classification module.The convolution module extracts the local features of EEG signals through spatial convolution and temporal convolution.The self-attention module extracts the global features of EEG signals through the spatial self-attention mechanism and the feature self-attention mechanism.The feature enhancement module further fuses local features and global features to extract features strongly related to categories.The classification module is responsible for predicting the probability of alcohol addiction.The experimental results show that the proposed model can effectively detect alcohol addiction,and the accuracy on the public data set can reach 100.00%,which is better than most of the current algorithms.

关 键 词：脑电图 酒精成瘾 自注意力机制 特征增强 

分 类 号：TN911.7[电子电信—通信与信息系统] TP183[电子电信—信息与通信工程]