摘要
脑磁信号中包含许多与精神疾病相关的生理信息,是判断神经系统出现各种异常活动的重要依据,对脑科学的研究具有十分重要的意义。为了提高正常人与精神分裂症患者的MEG数据的分类精度,文中提出了一种基于小波包基与能量熵的脑磁特征提取与识别的方法。该方法首先将经PCA降维后的MEG数据进行小波包分解,并结合小波熵从小波包库中选择最优小波包基,然后对选出的最优基所对应的小波系数进行幅度调制求取能量熵,并求取包络能量熵的统计特性构成分类特征向量,最后输入到SVM分类器,实现MEG数据的自动分类。实验结果表明,该方法的分类准确度可达到97.586 8%。说明文中提出的特征提取方法能够有效提取脑磁信号的特征,提高分类精度;也将为精神分裂症的诊断和严重程度的评估提供选择依据。
MEG signals associated with many physiological information related to mental illness is an important basis for judging abnormal nervous system activity and has great significance for the study of brain science. In order to improve accuracy of normal and schizophreni- a＇ s MEG signals,a new method of MEG classification with feature extraction is proposed based on wavelet packet and energy entropy. First,it reduces the dimension of the raw signals by Principal Component Analysis （PCA） and decomposes preprocessed signal by Wave- let Packet Decomposition （WPD）. Then, selects the best basis of wavelet packets from a wavelet packet library according to the wavelet packet entropy,afterwards calculating the energy entropy of envelope that acquired by Amplitude Modulation （AM） of the best basis wavelet coefficients. Moreover,the eigenvector is obtained by calculating the statistical features of energy entropy. Finally, the feature vec- tors are put into a Support Vector Machine （SVM） to realize automatic classification of MEG. Experiment shows that the proposed math- od could achieve a great classification aceuracy of 97. 586 8%, which indicates that the feature extraction method in this paper can effec- tively extract the characteristic of MEG and improve the classification accuracy. It provides evidence for the treatment and severity assess- ment of schizophrenia.
出处
《计算机技术与发展》
2016年第6期127-132,共6页
Computer Technology and Development
基金
国家自然科学基金资助项目(61271334)
关键词
脑磁信号
小波包分解
幅度调制
能量熵
支持向量机
MEG
wavelet packet decomposition
amplitude modulation
energy entropy
support vector machine
