课题目的是研究面向特征抽取鉴别EEG信号是否具有癫痫的表征,从而实现EEG信号的自动检测与分析。具体方法是结合临床EEG分析的先验知识,模拟结构匹配性稀疏表示的层次处理机制来实现对EEG信号的结构自适应稀疏分解。结果发现:匹配追踪迭代选择的Gabor字典原子能够匹配EEG信号的内在结构,并具有显式的形态结构参数如位置、尺度、幅度等。由此可得出结论:基于EEG信号形态结构基础建立的过完备原子库,使得稀疏分解获取的信号时频结构参数同人工视觉分析标准建立了直接联系。应用这些时频结构参数与先验参数进行比对可直接判定是否为特征波形。
Objective: Distinguishes the EEG signal face the feature extraction whether to have epilepsy's attribute, thus realizes EEG the signal automatic detection and the analysis. Methods: Clinical EEG analysis with a priori knowledge of the structure matching simulated sparse, said the level of processing mechanism to achieve the structure of the EEG signal adaptive sparse decomposition. Results: The matching pursuit iterations selected Gabor dictionary to match the atomic internal structure of EEG signals and has the explicit form of structural parameters such as position, scale and amplitude. Conclusion: Based on the EEG signal morphology-based database established over-complete atoms, making sparse decomposition to obtain time-frequency analysis of structural parameters of the standard with artificial vision to establish a direct link to apply these time-frequency structural parameters and a priori parameters can be compared directly to determine whether the characteristics of the waveform.