使用高階統計法則實現相位鍵移調變訊號分類作業

Abstract

綜觀軍事通訊發展，諸如頻譜管理、軍事監控、訊號情報蒐集、電子戰運用，大多通聯訊號相關參數無法事前獲得，因此如何在數據訊號偵收系統截獲敵方之未知通聯訊號時，能夠即時完成訊號辨識、解調，快速獲得重要情報，變成一個重要課題，其中訊號類型自動辨識及分類技術，成為接收端於射頻訊號截收至基頻訊號解調過程之必要執行項目，通聯訊號自動分類技術依基底區分2類，分別為假設判別(decision-theoretic)基底及特徵識別(pattern-recognition)基底，在本篇論文中，主要採用屬於特徵識別為基底之高階統計法則(moments以及cumulants)，實現常用於軍事通聯之相位鍵移調變(M-ary PSK)類型自動分類技術，因此本論文首先分析並模擬實現2階、4階、6階及8階等高階統計法則在所遭遇條件下(如理想情形、頻率偏移、相位偏移及時間偏移)展現特性及影響，接著就些許現已發表相關論文，模擬在多重路徑傳輸通道及可加性高斯白雜訊(AWGN)條件下之效能分析及遭遇困境，最後引出本論文針對困境所提出相位鍵移調變訊號分類法則，敘述如下: 1.使用4階moments進行通道脈衝響應參數估測及通道補償作業。 2.採用階級式(hierarchical)架構進行訊號分類，分別使用經過通道補償後之2階cumulants之特徵純量(feature scale)及4階cumulants之特徵純量。 數值分析結果顯示，當相位鍵移調變訊號於多重路徑傳輸條件下，本篇論文所提法則較其他採用4階cumulants之特徵向量(feature vector)或6階cumulants之特徵向量具備較佳分類效能，可有效運用於相位鍵移調變訊號自動分類應用上。

While investigating the development of military communications such as spectrum management, signal intelligence (SIGINT) collection and electronic warfare, most signal-related parameters are unavailable in advance. It has become an important topic to recognize the unknown signals and then extract the SIGINT from the signals quickly when these signals emitted from the enemy are intercepted by any signals-collection system. Automatic modulation recognition and classification has become a necessary step between signal detection and signal demodulation. Automatic modulation classification could be divided into two subgroups, decision-theoretic based and pattern-recognition based. In the thesis, higher-order statistics, moments and cumulants, is adopted to classify M-ary PSK (M-PSK) signals widely used by military communications. Firstly, second-order cumulants to eight-order cumulants are implemented and analyzed under the scenarios of the ideal situation, frequency offset, phase rotation, phase noise, timing offset and AWGN. Secondly, multipath fading channel is unavoidable in practice. So that performance comparison and drawback of some published articles are also implemented and discussed in detail while the transmitted signal is corrupted by multipath fading and AWGN. Finally, the proposed method in the thesis to recognize M-PSK signals is presented as following statements: (1)Blind channel estimation and compensation are carried out via fourth-order moments of the received signal without any a priori information. (2)The structure of proposed modulation classification method is hierarchical. Normalized second-order cumulants and normalized fourth-order cumulants after blind channel compensation are adopted to classify M-PSK signals. After the simulation and numerical analysis, the proposed method could get better performance than some other methods which adopt the feature vector of fourth-order cumulants or sixth-order cumulants as the feature extraction in multipath fading channel. The proposed method could be used in M-PSK signals classification efficiently.