凌宇 杜玉晓 李向欢
（广东工业大学，广东 广州 510006）
摘要：随着癫痫脑电信号自动检测算法研究地不断深入，需要处理的特征维度也不断增加，且冗余特征增大了算法的复杂度，导致算法性能下降。为此，提出一种基于F-Score特征选择的癫痫脑电信号识别方法。首先，从原始癫痫脑电信号数据集中提取特征，并计算每个特征的F-Score统计值；然后，根据分类模型的分类准确率，通过序列前向搜索方法，选择最优特征集；最后，利用支持向量机和逻辑回归分类模型进行实验，并与传统的特征降维方法PCA进行对比。实验结果表明，本文方法可有效降低特征矩阵的维数，提高算法运算效率。
关键词：F-Score；PCA；特征提取；特征选择；癫痫脑电信号识别
中图分类号：R742.1 文献标志码：A 文章编号：1674-2605(2023)05-0009-06
DOI：10.3969/j.issn.1674-2605.2023.05.009
Epileptic EEG Signal Recognition Method Based on F-Score Feature Selection
LING Yu DU Yuxiao LI Xianghuan
(Guangdong University of Technology, Guangzhou 510006, China)
Abstract: With the continuous deepening of research on automatic detection algorithms for epileptic EEG signals, the number of feature dimensions to be processed continues to increase, and redundant features increase the complexity of the algorithm, leading to a decrease in algorithm performance. To this end, a method for epileptic EEG signal recognition based on F-Score feature selection is proposed. Firstly, extract features from the original epileptic EEG signal dataset and calculate the F-Score statistical value for each feature; Then, based on the classification accuracy of the classification model, the optimal feature set is selected through a sequence forward search method; Finally, experiments were conducted using support vector machines and logistic regression classification models, and compared with the traditional feature dimensionality reduction method PCA. The experimental results show that the proposed method can effectively reduce the dimensionality of the feature matrix and improve the computational efficiency of the algorithm.
Keywords: F-Score; PCA; feature extraction; feature selection; epileptic EEG signal recognition
