高階頻譜分析應用於複雜頻譜特性之探討

Abstract

高階頻譜分析應用於複雜頻譜特性之探討 研究生: 劉怡劭 指導教授: 羅佩禎 老師 國立交通大學控制工程研究所 摘要 過去幾十年來,傳統的傅利葉轉換廣泛的應用於特性穩定訊號之分析. 但對於EUS-EMG (External Urethral Sphincter Electromyography) 及 EEG (Electroencephalograph) 等訊號而言,其所含之複雜頻譜特性卻無 法很有效的經由傅利葉分析來顯現出來.因此,在本論文中,我們主要研究 目的在於利用高階頻譜分析來萃取生醫訊號的複雜頻譜特性. 在 本篇論文中,我們首先利用傅利葉轉換來分析處於各種生理狀態下之 EUS-EMG 訊號以及EEG 訊號,接著以高階頻譜分析來萃取傅利葉轉換所無 法簡易清晰顯現的相位特性.此外,對於EUS-EMG訊號在各種狀態底下彼此 之關係以及各頻道EEG 訊號間之交互特性問題,也都是本篇論文研究的重 點. 根據本論文研究分析之結果,我們展現了高階頻譜分析之特性 與有別於 傳統傅利葉分析之優點.在將來的研究發展方面,我們將著 重在運用各種高 階頻譜分析之參數來拓展其於生醫訊號處理方面之 應用. Bispectral Analysis for Characterizing Signals with Complicated Spectral FeaturesStudent:Yi-Shao Liu Advisor:Pei-chen Lo Institute of Control Engineering National Chiao Tung University Abstract Conventional Fourier Transformation(FT) analysis is widely used in the areaof analyzing stationary signals. Demands can not be satisfied by only FT analysis since the EUS-EMG (External Urethral Sphincter Electromyography) andEEG (Electroencephalography) analyzed in this thesis are with complicatedspectral features. Bispectral analysis is then used to characterize complicated spectral properties in the bio- medical signals. In this thesis, the EUS-EMG signals were analyezd first by FT analysis and,then by bispectral and cross- bispectral analysis in order to remedy the phase-blind property of FT analysis. Furthermore, we investigate the correlationsbetween EUS-EMG signals under different activities. Secondly, EEG signals were analyzed by bispectral and cross- bispectralanalysis. The QPC features of EEG signals under different physiological stateswere characterized by the bispectral analysis. The spatial correlation of brain electrical activities were explored by cross-bispectral analysis. The results of this thesis depict the superiosity of bispectral and cross-bispectral analysis over conventional FT analysis. For further researches,more parameters can be derived from the bispectrum to extend its applicationsin biomedical signal processing area.