| 引用本文: | 何 爽,黄 鑫,林思睿,徐澜菲,王莹莹,邹任玲,李 丹,胡秀枋.基于神经网络的运动想象分类研究[J].软件工程,2022,25(8):6-10.【点击复制】 |
|
| 摘 要: 近年来，神经网络的模型不断得到完善，神经网络在运动想象分类任务中的应用越来越广泛，分类准确率不断提高。本文主要对传统的机器学习算法进行介绍与总结，在此基础上对深度学习网络模型的原理及应用进行了概括，主要分析卷积神经网络、生成对抗网络和胶囊网络这几种网络模型的优缺点及应用，并对多种网络模型组合分类或将单一网络模型中的多种特征进行组合分类的发展趋势进行展望，提出目前运动想象分类任务面临的问题及发展趋势。 |
| 关键词: 神经网络；运动想象；脑机接口；分类算法 |
|
中图分类号: TP391.41
文献标识码: A
|
| 基金项目: 上海市科技创新行动计划产学研医合作领域项目“基于电刺激的步态模拟下肢康复训练仪研制与应用”(21S31906000). |
|
| Research on Classification of Motor Imagery based on Neural Network |
|
HE Shuang, HUANG Xin, LIN Sirui, XU Lanfei, WANG Yingying, ZOU Renling, LI Dan, HU Xiufang
|
(School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China )
heshuangedu@163.com; hx2946474935@163.com; 1023617140@qq.com; 1046758980@qq.com; 3294576338@qq.com; zourenling@163.com; lidan0734454454@163.com; huxiufang1965@163.com
|
| Abstract: In recent years, with the constant improvement of neural network model, neural network is more and more widely used in the task of motor imagery classification, and the classification accuracy has been continuously improved. This paper mainly proposes to introduce and summarize the traditional machine learning algorithm. On this basis, it summarizes the principle and application of deep learning network model, and mainly analyzes the advantages, disadvantages and applications of convolutional neural network, generative adversarial network and capsule network. The development trend of combined classification of multiple network models or combined classification of multiple features in a single network model is prospected. The problems of the current classification task of motion phenomena and its development trend are presented. |
| Keywords: neural network; motor imagery; brain computer interface; classification algorithm |
