本文選題：深度學(xué)習(xí) + 卷積神經(jīng)網(wǎng)絡(luò)��； 參考：《哈爾濱工業(yè)大學(xué)》2017年碩士論文

【摘要】：航天測控在軍事領(lǐng)域的應(yīng)用非常廣泛,衛(wèi)星對抗成為空間信息偵查與對抗的主要領(lǐng)域。在這個基礎(chǔ)上,研究航天測控鏈路的干擾感知、自動識別具有非常重要意義。經(jīng)過多年的發(fā)展,干擾信號的感知和識別已經(jīng)取得了很多的成果,但是大多數(shù)的方法只能針對特定的干擾類型進行檢測,具有很大的局限性。同時,大多數(shù)的干擾類型識別的特征提取使用的是傳統(tǒng)的模式識別方法,需要人工進行干擾信號的特征的提取,這需要消耗很大的工作量。本文從能量檢測算法和深度學(xué)習(xí)網(wǎng)絡(luò)著手,研究干擾信號的檢測、特征識別和參數(shù)估計的機理及實現(xiàn)方法。本文首先對有涉及到的5種需要感知識別的干擾信號進行了分析說明,5種干擾信號包括音頻干擾、同頻帶窄帶干擾、掃頻干擾、矩形脈沖干擾以及擴頻干擾。之后,通過對能量檢測算法和恒虛警檢測算法的研究,成功使用能量檢測算法對干擾信號進行了存在性的檢測,實驗表明干擾信號在干噪比較高的環(huán)境下取得了很好的成果,但是在干噪比較低的環(huán)境下,能量檢測算法的性能急劇下降。因此,本文提出了使用深度學(xué)習(xí)網(wǎng)絡(luò)和能量檢測算法對干擾信號進行聯(lián)合檢測的方法。實驗結(jié)果表明,基于卷積神經(jīng)網(wǎng)絡(luò)的干擾檢測算法在干噪比很低的環(huán)境下檢測性能優(yōu)于能量檢測算法。然后使用了深度學(xué)習(xí)網(wǎng)絡(luò)對5種干擾信號進行了特征的提取,使用多維尺度分析方法對提取出的特征信息進行了分析,結(jié)果表明提取出的特征具有明顯的可分性與魯棒性,接著將提取出了干擾信號特征送入Softmax分類器進行分類。實驗結(jié)果該種分類方法在干噪比為-5dB~15dB時對單一干擾信號的分類正確率幾乎達到了100%,而并存干擾的分類正確率達到了99%以上。最后,利用四階統(tǒng)計特性代替原能量檢測器的平方特性,在感知到對方干擾衛(wèi)星通信后,利用“窗”的思想,采用第二類切比雪夫濾波器估計窄帶干擾信號的中心頻率和寬帶干擾信號的帶寬。實驗結(jié)果,在干噪比為-5dB~15d B的環(huán)境下干擾信號參數(shù)估計的準(zhǔn)確率在96%以上。
[Abstract]:Space TT & C is widely used in military field, satellite countermeasure becomes the main field of space information detection and countermeasure.On this basis, it is very important to study the interference perception and automatic recognition of space TT & C link.After years of development, many achievements have been made in the perception and recognition of interference signals, but most of the methods can only be detected for specific types of interference, which has great limitations.At the same time, most of the feature extraction of interference type recognition uses the traditional pattern recognition method, which needs to extract the feature of interference signal manually, which needs a lot of work.In this paper, energy detection algorithm and depth learning network are used to study the mechanism and implementation of interference signal detection, feature identification and parameter estimation.In this paper, firstly, five kinds of interference signals which need to be sensed and identified are analyzed. The five kinds of interference signals include audio interference, narrowband interference in the same frequency band, sweep interference, rectangular pulse interference and spread spectrum interference.Then, through the research of the energy detection algorithm and the constant false alarm detection algorithm, the existence of the interference signal is detected successfully by using the energy detection algorithm. The experiment shows that the interference signal has achieved good results in the environment of high dry noise.But in the environment of low dry noise, the performance of the energy detection algorithm drops sharply.Therefore, a method of joint detection of interference signals using depth learning network and energy detection algorithm is proposed.Experimental results show that the detection performance of the interference detection algorithm based on convolution neural network is better than that of the energy detection algorithm under the condition of low dry-noise ratio.Then, the features of five kinds of interference signals are extracted by using the deep learning network, and the feature information is analyzed by using multidimensional scale analysis method. The results show that the extracted features have obvious separability and robustness.Then the feature of the interference signal is extracted and sent into the Softmax classifier for classification.Experimental results show that the classification accuracy of single interference signal is almost 100 when the dry noise ratio is -5 dB, and the classification accuracy rate with interference is more than 99%.Finally, the fourth-order statistical characteristic is used instead of the square characteristic of the original energy detector. After sensing the interference of the other side to the satellite communication, the idea of "window" is used.The second kind of Chebyshev filter is used to estimate the center frequency of narrowband interference signal and the bandwidth of wideband interference signal.The experimental results show that the estimation accuracy of interference signal parameters is over 96% under the environment of -5 dB / 15dB.
【學(xué)位授予單位】：哈爾濱工業(yè)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TN97;V556

【參考文獻】

相關(guān)期刊論文 前10條

1 黃瑩;郝新紅;孔志杰;張彪;;基于熵特征的調(diào)頻引信目標(biāo)與干擾信號識別[J];兵工學(xué)報;2017年02期

2 周新;姚富強;牛英滔;;一種新的調(diào)頻干擾信號檢測算法[J];軍事通信技術(shù);2016年04期

3 Xiaokai Liu;Rong Li;Chenglin Zhao;Pengbiao Wang;;Robust signal recognition algorithm based on machine learning in heterogeneous networks[J];Journal of Systems Engineering and Electronics;2016年02期

4 譚文學(xué);Zhao Chunjiang;Wu Huarui;;CNN intelligent early warning for apple skin lesion image acquired by infrared video sensors[J];High Technology Letters;2016年01期

5 沈宇;張風(fēng)國;陳奇東;甄衛(wèi)民;;基于循環(huán)譜法的GNSS RF干擾信號檢測[J];全球定位系統(tǒng);2015年06期

6 齊佩漢;李贊;司江勃;熊天意;;A two-stage spectrum sensing scheme based on energy detection and a novel multitaper method[J];Chinese Physics B;2015年04期

7 王道明;魯昌華;蔣薇薇;肖明霞;李必然;;基于粒子群算法的決策樹SVM多分類方法研究[J];電子測量與儀器學(xué)報;2015年04期

8 溫媛媛;魏佳圓;陳豪;;一種新型的干擾識別與檢測算法[J];空間電子技術(shù);2015年01期

9 黃曉飛;徐池;;衛(wèi)星通信干擾樣式研究[J];航天電子對抗;2011年06期

10 鄭鵬;張鑫;劉鋒;;直擴信號載頻、碼元速率循環(huán)譜估計改進算法[J];電子測量技術(shù);2011年09期

相關(guān)碩士學(xué)位論文 前9條

1 鐘天宇;基于噪聲魯棒性的通信信號調(diào)制模式識別方法研究[D];哈爾濱工業(yè)大學(xué);2016年

2 宋希珍;海雜波幅度分布參數(shù)矩估計和恒虛警檢測[D];西安電子科技大學(xué);2014年

3 鄧文林;大動態(tài)信噪比下的多體制通信信號自動調(diào)制識別算法研究[D];哈爾濱工業(yè)大學(xué);2014年

4 陳先昌;基于卷積神經(jīng)網(wǎng)絡(luò)的深度學(xué)習(xí)算法與應(yīng)用研究[D];浙江工商大學(xué);2014年

5 李林甫;干擾識別技術(shù)研究與實現(xiàn)[D];西安電子科技大學(xué);2014年

6 許可;卷積神經(jīng)網(wǎng)絡(luò)在圖像識別上的應(yīng)用的研究[D];浙江大學(xué);2012年

7 呂再興;通信對抗中的干擾檢測算法研究[D];電子科技大學(xué);2011年

8 趙曉迪;基于譜分析的通信信號調(diào)制識別與參數(shù)估計研究[D];西南交通大學(xué);2010年

9 孫永亮;基于高階累積量的頻譜感知技術(shù)仿真分析[D];哈爾濱工業(yè)大學(xué);2008年

，

本文編號：1753761