摘要
近年来,随着脑机接口(Brain-Computer Interface,BCI)技术的进一步发展,对特征提取技术的鲁棒性的需求也持续增加。深度学习(Deep Learning,DL)作为多层次的神经网络模型具有从高维数据中进行特征提取并从分层表示中学习的能力,在分类识别任务领域中的表现优于手工选择特征的传统机器学习方法。深度学习模型可以自动学习高维的EEG数据集从而提取有效特征,因此基于深度学习的脑机接口成为该领域新的研究趋势。卷积神经网络(Convolution Neural Network,CNN)、深度信念网络(Deep Belief Network,DBN)和递归神经网络(Recurrent Neural Network,RNN)是深度学习中对脑电信号进行分析的三大主流算法。主要介绍了这三大主流深度学习算法的基本原理。为了探索能更好契合脑电数据特点的分类模型,还探讨了它们在BCI中集成其他方法的实际运用。
In recent years,with the further development of BCI(Brain-Computer Interface)technology,there has been a growing demand for robustness of feature extraction techniques.DL(Deep Learning),as a multi-layered neural network model with the ability to extract features from high-dimensional data and learn from hierarchical representations,has outperformed conventional machine learning methods that rely on handcrafted features in classification and recognition tasks.DL models can automatically learn high-dimensional EEG datasets and thus extract effective features,making DL-based BCIs a new research trend in the field.CNNs(Convolutional Neural Networks),DBNs(Deep Belief Networks),and RNNs(Recurrent Neural Networks)are the three mainstream algorithms for analyzing EEG signals in DL.This paper introduces the basic principles of these three mainstream deep learning algorithms.In order to explore classification models that better fit the characteristics of EEG data,this paper also discusses their practical use in integrating other methods in BCI.
作者
陈茂洲
刘化东
CHEN Maozhou;LIU Huadong(Kunming University of Science and Technology,Kunming Yunnan 650500,China)
出处
《通信技术》
2023年第6期673-681,共9页
Communications Technology
关键词
脑机接口
深度学习
特征提取
脑电
brain computer interface
deep learning
feature extraction
EEG
