【作者】 王立东；

【导师】 张晓林；

【作者基本信息】 哈尔滨工程大学 ， 通信与信息系统， 2011， 硕士

【摘要】 随着通信体制及信号调制样式的多样化和复杂化,无线通信环境越来越复杂,通信信号在时域、频域、空域上交织重叠,使信号调制样式的识别难度越来越大,这也使通信信号的调制方式识别研究成为军、民领域极为关注的课题。本文在借鉴已有知识基础上,提出了基于变形光滑支持向量机通信信号调制方式识别的新方法,对混合通信信号中各类别信号的调制方式进行了识别研究和仿真分析。首先分析了7种常用数字信号的调制原理,对目前常用的特征提取和载波频率估计方法进行了研究,提取6类特征作为分类器识别的特征向量。然后针对接收信号多为混叠信号、先验知识少、分离困难等特点,引入独立分量分析的方法,从若干独立信源混合后的观察信号中分解出各个独立的信号。最后,对支持向量机的相关理论进行了研究。针对传统支持向量机二次规划问题的目标函数具有二阶不光滑、不可微的特点,提出了一类变形光滑支持向量机模型(SCHKS-SSVM),采用CHKS函数作为光滑函数,并用Newton-Armijo算法来训练该模型,通过批处理训练提高训练速度,节省了大量存储空间。实验仿真表明,该方法可以有效求解高维、大规模的分类问题。更多还原

【Abstract】 With the diversification and the complication of communication system and signal modulation pattern, the wireless communication environment is more and more complicated, the communication signals overlap in the time domain、the frequency domain and the airspace, it makes signal recognition more difficult and makes the research of the communication signals modulation identification become attention topic in civil and military field. Referencing to the existing knowledge, this paper puts forward a new identification method based on the deformation smooth support vector machine. Simulation proves that this method can recognize the modulation mode of hybrid communication signals.First, this paper analyses the modulation principle of seven common digital signals. Given the study of the usual feature extraction and the carrier frequency estimation, choose six characteristics to be recognition characteristic vector. And then, according to the receive signal for aliasing signal、prior knowledge less、separation difficult and so on, introducing to independent component analysis method. It separates the independent composition from the observed signals which are mixed by several independent sources. Finally, researchs the theory about the support vector machine (SVM). The second order objective function of traditional support vector machine quadratic programming problem is not smooth、not micro. So it introduces a smooth deformation support vector machine (SCHKS-SSVM) mode, which uses CHKS function as smooth function. The model is trained by Newton-Armijo algorithm. Beacause it can improves the training speed through the batch train, meanwhile saving a lot of storage space. Experiments show that the proposed method can effectively solve high-dimension、large-scale classification problem.更多还原