摘要
为了有效识别癫痫脑电信号,提出了一种适合于非平稳脑电信号的特征提取方法。以临床采集的包含癫痫发作期的5组500个EEG公共数据为样本,选择了具有任意多分辨分解特性的小波包变换,对信号进行多尺度分解,并提取了各级节点的小波包系数。将小波包系数能量作为特征值,构建了特征向量并输入到BP神经网络分类器中进行自动识别。实验结果表明,该算法的识别率达到了91.5%。
A method of feature extraction in non-stable signals is put forward to improve the correct classification rates of epileptic EEG. The Samples are composed of five hundred EEG Public datum which include the Period of epileptic seizures. The authors select the wavelet packets that have the trait of arbitrary distinction and decomposition. Character vectors which reflect different state of EEG signals are extracted from different frequency segments with the technology of wavelet packet decomposition, and taking them input neural network as samples to establish the model of BP neural network. Extensive experimental results demonstrate that the classification accuracy of the proposed feature extraction method for experiment EEG signals reach 91.5%.
出处
《电子测量技术》
2009年第10期36-39,共4页
Electronic Measurement Technology
关键词
癫痫脑电
小波包
特征提取
BP神经网络
epileptic EEG
wavelet package
feature extraction
BP neural network
