基于多时窗共空间模式的HMM运动想象脑电识别

Motor Imagination EEG Recognition Based on a Hidden Markov Model of Multi-Time Window Common Spatial Pattern

摘要:

运动想象脑电具有识别效果不佳及复杂时序信号建模困难的问题；提出一种基于多时窗共空间模式的隐马尔可夫模型运动想象脑电识别方法,首先将运动想象脑电划分为多个短时窗信号,然后使用共空间模式提取特征序列,以滤除脑电通道间的冗余信息,最后采用前向-后相算法与Viterbi算法求解隐马尔可夫模型并完成分类识别；将本文方法在公开运动想象脑电数据集上进行实验，得到77.17%的分类正确率，相较隐马尔可夫模型算法提升了5.74%，验证了所提方法的有效性。

Abstract:

Aiming at the poor performance of motor imagery electroencephalogram recognition and the difficulty in modeling complex signals, this paper proposes a motor imagery electroencephalogram recognition method based on a hidden Markov model of multi-time window common spatial pattern. Firstly, the motor imagery electroencephalogram was divided into different time window signals, and then the common spatial pattern was used to extract feature sequences, which can filter out redundant information between electroencephalogram channels. Finally, the forward-backward algorithm and the Viterbi algorithm were used to solve the hidden Markov model and complete classification recognition. The method proposed in this paper was validated on a publicly available motor imagery electroencephalogram dataset, and the classification accuracy rate is 77.17%, which is 5.74% higher than that of the Hidden Markov model algorithm, verifying the effectiveness of the proposed method.