本研究提出一個利用"希爾伯-黃"轉換(Hilbert-Huang Transformation; HHT)對污染ST/MST 雷達之飛機及其它暫態雜訊進行濾波。ST/MST 雷達主要求取大氣回波及其風場，但常受到地表物及暫態回波之干擾以致資料無法加以利用。地表物的回波因為位於頻譜中零頻附近，故易於利用訊號分析的方法去除掉其成份，但對於飛機相關訊號則常無法克服其干擾，主要是過去利用傅立葉分析，此種方法或其衍生出來的技術對於非線性且短暫出現的訊號無法處理。在HHT 方法出現之後，不需要假設訊號源之數學特性，依照其即時頻率區分成若干個Intrinsic mode function，再進行希爾伯轉換後，可得一個隨著時間及頻率分佈的能量譜，且同成份之能量不會重複出現在其它的時頻區域，此提供了我們一個濾波的新想法，利用 HHT 來去除掉最難去除的飛機雜訊。We propose a new filter technique based on the Hilbert-Huang transform (HHT) to remove airplane and other transitory echoes in the ST/MST radar measurements. ST/MST radars congregate information regarding the three dimensional atmospheric winds. Apart from this important information, they also receive unwanted ground (due to stationery targets such as tall buildings and trees) and intermittent clutter (due to non-stationery targets such as airplanes, and birds) that needs to be removed before a spectral analysis. Ground clutter always appears around zero frequency and can hence be detected and filtered if it does not mix with the clear-air signal. Though traditional data analysis method such as Fourier analysis has been proved to be efficient in removing unwanted signals of the type that associated with ground clutter, these methods are of less efficient in analyzing non-linear and non-stationery signals. Besides, wavelet and the Wagner–Ville distribution (Heisenberg wavelet) analysis are designed for linear but non-stationary data. On the other hand, with the HHT technique, time series decomposed into several ‘intrinsic mode functions (IMF)’ yield instantaneous frequencies of themselves. The final presentation of the results is an energy-frequency-time distribution, designated as the Hilbert spectrum. From the result, one can find a general separation of the data into locally non-overlapping time scale components. In some components, the signals are intermittent, then the neighboring components might contain oscillations of the same scale, but signals of the same time scale would never occur at the same locations in two different IMF components. This allows us to remove the unwanted data according to its corresponding instantaneous frequency that is larger than a presumed threshold value. Future works to be done to improve the performance of this filter are discussed. 研究期間 : 9808 ~ 9907
