基于聚类理论的任意星座图MQAM信号调制方式识别算法研究

项目名称： 基于聚类理论的任意星座图MQAM信号调制方式识别算法研究

项目编号： No.U1204604

项目类型： 联合基金项目

立项/批准年度： 2013

项目学科： 电子学与信息系统

项目作者： 孙钢灿

作者单位： 郑州大学

项目金额： 30万元

中文摘要： MQAM信号调制阶数范围很大，调制星座图的样式种类多，对MQAM信号进行调制识别时，难以提前获取待识别信号的标签样本集，这是调制识别研究领域的难点。在模式识别理论中针对类似问题用非监督的模式识别方法或基于聚类的方法来解决，因此，本项目拟研究基于聚类理论的任意阶数任意星座图MQAM信号的调制方式识别问题，分析将聚类理论引入MQAM信号的调制识别时可能面临的难题，探讨问题产生的原因，研究用聚类理论解决星座图重构时的规律，在此基础上进行总结归纳给出问题的解决方法，最终给出基于聚类的调制方式识别的完整理论。拟研究的主要问题和解决思路有：1、初始聚类中心的确定，拟采用减法聚类算法，该方法速度快，但需研究如何设定聚类半径和聚类结束判决条件。2、二次聚类算法，研究如何通过均值聚类进一步纠正减法聚类出现的误差。3、探讨算法运算复杂度问题，复杂度跟样本的数量有关，找到保证一定聚类精度下所需样本数量的最小值。

中文关键词： MQAM信号；调制识别；星座图；聚类算法；

英文摘要： The range of modulation order of QAM signals can be from a little number to a very large number. The QAM signals have many different type constellations. In practical works, we may encounter a new constellation that we have never met before. This makes modulation recognition of QAM signal becomes a difficult task. This problem can be solved by unsupervised pattern recognizing method or clustering method. This program mainly reseachs the modulation recognition technique of MQAM signals with any constellations, and search for the method of reconstruction the constellations of MQAM signals using clustering algorithm. Firstly, this program analyses on the problems caused by clustering method. Then, we study the reasons of these problems and the principles of MQAM constellation reconstruction using clustering method. Finaly, we give the whole theory of modulation recognition based on clustering method.Key tasks are listed as follows: First, How can we get the initial centroid? An subtractive clustering may be use to solve this problem. But we will study the problems that input signals of different order modulation types require different radius values and how to to make a decision whether the procedure of clustering is over. Second, after subtractive clustering outputs the initial centroid, we use C-means or fussy C

英文关键词： MQAM signals；Modulation classification；Constellation diagram；Clustering Algorithm；