由声纳所量测到之原始氐与声学信号,若圱时.pdfVIP

小波封包為基底之希爾伯特轉換應用 於信號特徵參數擷取法研究 1 2 杜筑奎 梁家銘 摘要 進行信號特徵參數擷取之工作，包括信號特性分析及特徵參數求取兩部份。首先利用小波轉換之多重解 析度特性，以小波封包分解之方式對信號作分頻分解，再針對各子空間在頻譜之特性，選擇適合之子空間作 特徵擷取，同時為突顯信號內之暫態分量，以與動態分量同時成為代表信號之特徵值，故利用希爾伯特轉換 於水下聲源信號之分析。本研究針對水下信號之特性，發展以小波封包為基底之希爾伯特轉換，並將之應用 於信號特徵參數之擷取，為提高大量信號辨識之效率，故將小波封包分解後之信號，使用自組織圖演算法， 以產生出代表各類別信號特性之樣版特徵參數，最後再以希爾伯特轉換將信號轉換至頻率領域，此方法對於 分析窄頻信號或暫態信號都有良好效果，可達成高解析度信號特徵快速擷取之要求。 Study of Wavelet Packets Based Hilbert Transform Applying on Signal Feature Parameters Extraction 1 2 Tu Chu-Kuei Jia-Ming Liang ABSTRACT The task of signal feature parameter extraction consists of signal characteristic analysis and feature parameter search. Take the advantage of multi-resolution property of wavelet packets, the signal is decomposed to several sub-base, and then select appropriate sub-base to do the feature extraction, the Hilbert transform is applied on the feature to obtain the time-frequency distribution of energy. Based on the characteristic of underwater signal this research develop a wavelet packages based Hilbert transform method and apply on extraction of feature parameters of signal. In order to improve the efficiency of recognition for large amount signal a Self-Organizing Map Algorithm is used to cluster features of same category and produce the templet feature parameters for that category, then the Hilbert transform is used which transfer feature of signal to frequency domain and obtain the fast and accurate result .The bandwidth of signal obtained by this approach is narrow, in the mean time, the resolution of sig