|Abstract: ||訊號子空間法是近代發展被動式陣列訊號處理中用來估測多目標物方位的一個重要的方法,它具有高解析度及高精確度的優點。然而此方法主要的缺點,乃是必須假設雜訊的空間相關度為已知,或是不具有相關性(白色雜訊)。這種假設之成立,在實際上的雷達或是聲納系統是非常罕見的。當量測訊號具有高相關度雜訊成分時,現今的各種訊號子空間方法之估測特性具有明顯惡化情形,因而失去了原來高解析度及高精確度之能力。 本計畫研究的目的是針對上述的缺失,對於訊號子空間法提出改良的方法,使其估測性能可在未知高相關度的雜訊下,亦能兼具高解析度和高精確度之優點。當雜訊和訊號之機率分布有差異時,高階統計量可以被用來分辨訊號或是雜訊。例如當訊號為高斯分布,而雜訊為非高斯分布時,感測器所接收到的訊號,其三階度積量只含有來自於雜訊之成分,再利用高階統計量之投影特性,可以估測雜訊的相關性。一旦雜訊之相關性被估測後,訊號子空間法即可以被應用於方位之估測。本計畫除了新理論的推導發展外,並討論電腦模擬結果,以應證理論之正確性與可行性。|
Signal-subspace method is one of many important methods, developed recently in passive array processing, for the direction-of-arrival (DOA) estimation of multiple targets. It has the advantages of high resolution and high accuracy in terms of estimation performance. However, the major drawback of this method is the key assumption that the spatial correlation of noise field should be known or the noise field is uncorrelated. This assumption is not valid in practical applications such as radar and sonar systems. When highly correlated noise is encountered in measurements, the performance of any existing signal-subspace method degrades significantly, resulting in the loss of high resolution and high accuracy capabilities. To get over the problem mentioned above, we propose a means for improvements on signal-subspace method so that the resultant method still possesses capabilities of high resolution and high accuracy for the unknown highly correlated noise field. When probability distributions of signal and noise are distinctive, higher-order statistics can often be used for distinguishing between them. For example, when Gaussian signals are embedded in nonGaussian noise field, the third-order cumulant of the received array output contains only the contribution from noise. Therefore, the second-order correlation structure of noise field can be estimated by the use of projection property of higher-order statistics, and signal-subspace method can then be used for DOA estimation. The objective of the proposed project is to develop a robust DOA estimation method, based on signal-subspace method, by applying the theory of higher-order statistic for radar and sonar applications. Besides the new theoretical developments, computer simulations were performed in order to verify the feasibility and validity of the proposed method.