Electroencephalogram (EEG) signals are widely used as an effective method for epilepsy analysis and diagnosis. For the establishment of an accurate and efficient epilepsy EEG identification system, it is very important to properly extract the features of EEG signals and select appropriate combination features. This paper proposes an automatic epileptic EEG identification method based on hybrid feature extraction. It uses temporal and frequency domain analysis, nonlinear analysis and one-dimensional local pattern recognition method to extract epileptic EEG features. Gradient energy operator and local speed pattern are proposed to better reflect typical feature in the active EEG signals measured during seizure-free intervals. The genetic algorithm is used to select the obtained hybrid features; then the AdaBoost classifier is used to classify epileptic EEG under various classification conditions. Classification results on the dataset developed by University of Bonn show that the proposed method can be used to classify normal EEG, interictal EEG and seizure activity with only a few features. Compared with related researches using the same dataset, the proposed method can obtain an equally satisfactory classification accuracy while the feature amount is reduced by 61–95%. In particular, the classification accuracy of the interictal and normal EEG can reach 99%.

中文翻译：

---

基于混合特征提取的癫痫脑电图识别

脑电图（EEG）信号被广泛用作癫痫分析和诊断的有效方法。为了建立准确、高效的癫痫脑电识别系统，正确提取脑电信号的特征并选择合适的组合特征非常重要。本文提出了一种基于混合特征提取的癫痫脑电图自动识别方法。它采用时域和频域分析、非线性分析和一维局部模式识别方法提取癫痫脑电特征。提出了梯度能量算子和局部速度模式，以更好地反映在无癫痫发作间隔期间测量的活动脑电图信号的典型特征。采用遗传算法对得到的混合特征进行选择；然后使用AdaBoost分类器对各种分类条件下的癫痫脑电图进行分类。波恩大学开发的数据集上的分类结果表明，所提出的方法可用于对正常脑电图、发作间期脑电图和癫痫发作活动进行分类，只有少数特征。与使用相同数据集的相关研究相比，所提出的方法在特征量减少61-95%的情况下可以获得同样令人满意的分类精度。特别是发作间期和正常脑电图的分类准确率可达99%。与使用相同数据集的相关研究相比，所提出的方法在特征量减少61-95%的情况下可以获得同样令人满意的分类精度。特别是发作间期和正常脑电图的分类准确率可达99%。与使用相同数据集的相关研究相比，所提出的方法在特征量减少61-95%的情况下可以获得同样令人满意的分类精度。特别是发作间期和正常脑电图的分类准确率可达99%。