基于多特征融合的调制模式识别研究

发布时间：2018-02-27 02:10

本文关键词： 模式识别多特征融合聚类算法有效性　出处：《西南科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：在现代通信系统中,调制信号的复杂程度越来越高,一些通信系统需要对多种调制信号进行识别,但目前算法多数缺乏通用性,即识别种类有限。另外在常用算法中,较多使用单一特征参数对信号进行识别,在噪声较大时,很难进行有效识别。本文针对调制信号在噪声环境中识别率不高的问题,通过特征参数提取和分类器的改进和优化,研究MASK\MFSK\MPSK\MQAM等13种典型数字信号调制方式在低信噪比环境中的识别率。在特征参数提取方法中,本文使用了归一化样本熵特征,该参数能表征数据的复杂度程度;针对单一的特征参数识别率不高的问题,给出了基于多特征的多特征融合算法,将多个特征参数融合为可供调制信号类间识别的二进制特征参数,再进行模式识别,主要用于MASK\MFSK\MPSK\MQAM四类经典调制类型的识别。在类间识别中,本文给出了基于星座图的AP-FCM联合聚类算法,并使用聚类有效性准则检测聚类性能,该算法能够自适应的重构星座图,以此对2PSK\4PSK\8PSK\16PSK\16QAM\32QAM\64QAM调制信号进行识别;对MASK调制信号通过基于幅度谱密度和样本熵特征的融合进行识别,识别种类分别为2ASK\4ASK\8ASK调制信号;MFSK调制信号通过频谱的谱峰个数和谱峰间距的标准偏差特征的融合进行对2FSK\4FSK\8FSK调制信号识别。最后,还给出了基于MATLAB GUI的AP-FCM聚类算法的仿真界面设计。本文通过对改进和优化的算法的仿真和实验,在信噪比较低时,该算法能准确识别调制类型,有较高的识别率,并使用AP和FCM联合算法自适性重构星座图,辨识度较高,从而证明了经过多特征融合算法后的模式识别的有效性和正确性。
[Abstract]:In modern communication systems, the complexity of modulation signals is becoming more and more complex. Some communication systems need to recognize many kinds of modulated signals, but most of the current algorithms are short of generality, that is, the types of recognition are limited. It is difficult to identify the signal effectively when the noise is high. In this paper, the feature parameter extraction and the improvement and optimization of the classifier are used to solve the problem that the recognition rate of the modulated signal in the noise environment is not high. The recognition rate of 13 typical digital signal modulation methods such as MASK\ MFSK\ MPSK\ MQAM in low SNR environment is studied. In the feature parameter extraction method, the normalized sample entropy feature is used, which can represent the complexity of the data. In order to solve the problem that the recognition rate of single feature parameter is not high, a multi-feature fusion algorithm based on multi-feature is presented. The multi-feature parameters are fused into binary feature parameters which can be recognized between classes of modulation signals, and then the pattern recognition is carried out. It is mainly used for the recognition of four kinds of classical modulation types of MASK\ MFSK\ MPSK\ MQAM. In the inter-cluster recognition, this paper presents the AP-FCM joint clustering algorithm based on constellation diagram, and uses the clustering validity criterion to detect the clustering performance. The algorithm can adaptively reconstruct the constellation diagram. In this way, 2PSK\ 4PSK\ 8PSK\ 16PSK\ 16QAM\ 32QAM\ 64QAM modulation signal is recognized, and the MASK modulation signal is identified by fusion based on amplitude spectral density and sample entropy. The recognition category is 2ASK\ 4ASK\ 8ASK modulation signal MFSK modulation signal is identified by the fusion of the number of spectral peaks and the standard deviation feature of spectral peak spacing. Finally, the 2FSK\ 4FSK\ 8FSK modulation signal is identified. The simulation interface design of AP-FCM clustering algorithm based on MATLAB GUI is also given. Through the simulation and experiment of the improved and optimized algorithm, when the SNR is low, the algorithm can accurately identify the modulation type and has a high recognition rate. The adaptive algorithm of AP and FCM is used to reconstruct the constellation diagram, and the recognition degree is high, which proves the validity and correctness of the pattern recognition after multi-feature fusion algorithm.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.3

【相似文献】