首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于组合支持向量机的水声目标智能识别研究
引用本文:胡桥,郝保安,吕林夏,陈亚林,孙起,钱建平.基于组合支持向量机的水声目标智能识别研究[J].应用声学,2009,28(6):421-430.
作者姓名:胡桥  郝保安  吕林夏  陈亚林  孙起  钱建平
作者单位:中国船舶重工集团公司第705研究所,西安,710075
基金项目:中国博士后科学基金资助项目,中国船舶重工集团公司第705研究所总工程师基金 
摘    要:为解决水声目标小样本模式识别问题,有效地提高复杂海洋环境中的识别精度,提出了一种基于经验模式分解(EMD)、特征距离评估技术(FDET)和组合支持向量机(CSVMs)的水声目标智能识别方法。首先,将滤波、Hilbert包络解调和EMD等信号处理方法对水声目标的辐射噪声信号进行预处理,提取7个包含原始信号和预处理信号的时域和频域统计特征的特征集。然后,通过FDET从原始特征集中选择出7个敏感特征集。最后,将7个敏感特征集输入到7个支持向量机分类器中,利用遗传算法对7个分类器的结果进行合并,构成CSVMs分类器,从而实现对水声目标的智能识别。将该方法应用于舰船等水声目标的识别中,研究结果表明,该方法的识别性能优于单一SVMs分类器:同时,经过FDET得到的敏感特征集能明显地提高识别精度。

关 键 词:经验模式分解  特征提取  特征选择  组合支持向量机  水声目标识别

Intelligent underwater-acoustic-target recognition based on combination support vector machine
HU Qiao,HAO Bao-an,LV Lin-xi,CHEN Ya-lin,SUN Qi and QIAN Jian-pin.Intelligent underwater-acoustic-target recognition based on combination support vector machine[J].Applied Acoustics,2009,28(6):421-430.
Authors:HU Qiao  HAO Bao-an  LV Lin-xi  CHEN Ya-lin  SUN Qi and QIAN Jian-pin
Institution:(The 705 Research Institute,China Shipbuilding Industry Corporation,Xi’an 710075)
Abstract:To solve the small-sample pattern recognition problem of underwater acoustic target and improve the classification accuracy in complicated oceanic environment, a novel intelligent target recognition method for underwater acoustic signals is proposed, based on the empirical mode decomposition (EMD), the feature distance evaluation technique (FDET) and the combination support vector machines (CSVMs). Firstly, some signal preprocessing techniques, like filtration, Hilbert envelope-demodulation and EMD are performed on the radiated noise of underwater targets to preprocess target characteristic information. Then, seven feature sets, including time- and frequency-domain statistical features of both the raw and the preprocessed signals, are extracted. Secondly, with FDET, seven salient feature sets are selected from the seven original feature sets, respectively. Finally, the seven salient feature sets are put into the corresponding single classifier based on support vector machines (SVMs), and then the recognizing results of seven classifiers are combined with genetic algorithms. So the CSVMs classifier is constructed to recognize the underwater target. As an utilization, this proposed method is applied to some ship recognition. Testing results show that this proposed method has a better classification performance compared to the signal classifier based on single SVMs, and that with the salient feature set via FDET, the classification accuracy is greatly increased.
Keywords:Empirical mode decomposition  Feature extraction  Feature selection  Combination support vector machines  Underwater acoustic target recognition
本文献已被 CNKI 万方数据 等数据库收录!
点击此处可从《应用声学》浏览原始摘要信息
点击此处可从《应用声学》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号