稀疏降噪自编码器在IR-BCI的应用研究

发布时间：2018-03-09 16:35

  本文选题：模拟阅读　切入点：脑-机接口　出处：《计算机工程与应用》2017年11期 　论文类型：期刊论文

【摘要】：针对脑-机接口的特征提取问题,提出了一种基于非监督学习的稀疏降噪自编码器,对刺激诱发的脑电信号进行自主学习,构建原始数据的深层特征表达。该编码器引用稀疏自编码神经网络,通过加入噪声,增强其学习的泛化能力,增加了神经网络的鲁棒性。首先对多导联信号进行重新拼接,输入稀疏降噪自编码器,得到原始数据的稀疏特征表达;然后,采用支持向量机将学习到的特征进行分类;最后,同直接使用最优单通道相对比。实验结果为:稀疏降噪自编码器的分类准确率要优于单通道,表明该方法能够更好地学习到特征,并提高了"模拟阅读"脑-机接口的识别正确率,为脑-机接口系统的特征提取和分类提供了新思路。
[Abstract]:To solve the problem of feature extraction of brain-computer interface, a sparse de-noising self-encoder based on unsupervised learning is proposed, which can be used for autonomous learning of stimulus-induced EEG signals. The encoder uses sparse self-coding neural network to enhance its learning generalization ability and enhance the robustness of neural network by adding noise. Firstly, the multi-lead signal is reassembled. Input sparse denoising self-encoder to obtain sparse feature representation of the original data; then, support vector machine will be used to classify the features learned; finally, The experimental results show that the classification accuracy of sparse noise reduction self-encoder is better than that of single channel. The recognition accuracy of "simulated reading" brain-computer interface is improved, which provides a new idea for feature extraction and classification of brain-computer interface system.
【作者单位】： 中南民族大学医学信息分析及肿瘤诊疗湖北省重点实验室;中南民族大学认知科学国家民委重点实验室;
【基金】：国家自然科学基金(No.91120017,No.81271659) 中央高校基本科研业务费资助项目(No.CZY13031)
【分类号】：TN762;TN911.7;TP18
，

本文编号：1589349