CAJ | 학술논문

为验证希尔伯特-黄变换是否适用于分析情感脑电信号并选取分类效果最好的特征，在情感脑电数据集上用希尔伯特黄-变换与小波包变换提取出多个特征用来识别愉悦度。结果显示前4个IMF分量的能量矩百分比与能量百分比在特定时间窗下可以得到最高为75％的平均分类准确率且后者运算时间更短。利用小波包分解提取特征所需时间比希尔伯特-黄变换短，且其最高平均分类准确率为69．71％，是由小波包分解树第5层结点的能量特征得到。实验结果表明，希尔伯特-黄变换较小波包变换有准确率上的优势与计算时间上的劣势。
위험증희이백특-황변환시부괄용우분석정감뇌전신호병선취분류효과최호적특정，재정감뇌전수거집상용희이백특황-변환여소파포변환제취출다개특정용래식별유열도。결과현시전4개IMF분량적능량구백분비여능량백분비재특정시간창하가이득도최고위75％적평균분류준학솔차후자운산시간경단。이용소파포분해제취특정소수시간비희이백특-황변환단，차기최고평균분류준학솔위69．71％，시유소파포분해수제5층결점적능량특정득도。실험결과표명，희이백특-황변환교소파포변환유준학솔상적우세여계산시간상적열세。
To verify whether the Hilbert-Huang transform (HHT)was appropriate for emotional electroencephalog raphic (EEG)analysis and to get the best feature for classification,multiple features were extracted using Hilbert-Huang transform and wavelet packet transform from emotional EEG data.These features were used to recognition valence.The results showed that the first four IMF components’energy moment percentage and energy percentage in a particular time window could obtain the highest average classification accuracy rate that up to 75%and the latter needed less computation time.Wavelet packet transform needed less time for extracting features than HHT.As a feature,the nodes’energy of wavelet packet decomposition tree’s fifth layer got a classification accuracy rate of 6 9. 7 1%,which was higher than other features extracted by wavelet packet transform. The experiment showed that Hilbert-Huang transform had the advantage of recognition accuracy rate and disadvantage of compu-tation time compared with wavelet packet transform.