| 基于C-LSTM模型的端到端多粒度运动想象脑电信号分析方法 |
| |
| 引用本文： | 李海峰,徐聪,马琳.基于C-LSTM模型的端到端多粒度运动想象脑电信号分析方法[J].信号处理,2018,34(8):883-890. |
| |
| 作者姓名： | 李海峰 徐聪 马琳 |
| |
| 作者单位： | 哈尔滨工业大学计算机科学与技术学院 |
| |
| 基金项目： | 国家自然科学基金项目（61671187）；深圳市基础研究项目（JCYJ20150929143955341）；语言语音教育部-微软重点实验室开放基金资助项目(HIT.KLOF.20150XX, HIT.KLOF.20160xx)
|
| |
| 摘 要： | 脑电信号（Electroencephalography, EEG）是人的大脑在不同状态下产生的生物电信号。运动想象脑电信号是其中较为典型的一类信号，广泛应用于脑机接口技术中。对运动想象脑电信号分析的研究由来已久，目前主要采用公共空间模式等特征提取方法，对于如何提取更加有效的脑电信号特征以及如何对时序信息进行建模仍然是需要解决的问题。因此，本文设计了基于C-LSTM（Convolutional-Long Short Term Memory）模型的端到端多粒度脑电分析方法。并利用空间信息以及小波脑网络方法进行了改进，在BCI2008数据集上，相较传统方法提高了近10%，到达了93.6%的识别率。
|
| 关 键 词： | 脑电信号 多粒度特征提取 卷积神经网络 长短时记忆网络 |
| 收稿时间： | 2018-04-04 |
End-to-End Multi-Granular Motor Imagery EEG Signal Analysis Method Based on C-LSTM Model |
| |
| Affiliation: | School of Computer Science and Technology, Harbin Institute of Technology |
| |
| Abstract: | Electroencephalography (EEG) is a bioelectric signal generated by the human brain in different states. Motor Imagery EEG signals are one of the most typical types of EEG signals and are widely used in brain-computer interface technology. The research on motor imagery EEG signal analysis has a long history. And the methods such as Common Spatial Pattern are mainly used to extract features in the signal. It is still a problem to be solved how to extract more effective EEG signal features and how to model the timing information. Therefore, an end-to-end multi-granular EEG analysis method based on C-LSTM (Convolutional-Long Short Term Memory) model is designed in this paper. Then spatial information and wavelet brain network are used to improve the model. Compared with the traditional method, the experiment result on BCI2008 data set has improved by nearly 10%, reaching a recognition rate of 93.6%. |
| |
| Keywords: | |
|
| 点击此处可从《信号处理》浏览原始摘要信息 |
|
点击此处可从《信号处理》下载全文 |
