核域多重分形谱感知的流识别与分类

发布时间：2018-04-16 20:13

本文选题：流 + 识别与分类　；参考：《计算机应用研究》2017年07期

【摘要】：目前关于流识别与分类的主流技术是基于统计学方法,其核心环节是提取有效的特征属性集,但这种方法的假设条件是,特征互不相关、数据也互不相关。正因为这种假设的不合理性,使得分类效果和识别性能有限,引入以数据相关性为核心的多重分形理论,从根本上摈弃独立假设的局限性与狭隘性,实现流的有效分类。为此,定义并论证流的分形谱,在此基础上推导流的估计谱,然后在定义的核域内基于灰色关联度进行估计谱分析,继而脱离特征提取过程实现流的分类识别。最后通过系列实验显示流的分形性和分形谱,并进行实际分类效果的纵向比较和横向比较。研究结果表明,基于多重分形理论的流分类识别方法,有效弥补了统计学方法所不可避免的独立假设缺陷,因此具有强大而高效的识别未知流的能力,也特别适合于动态多变的在线识别。
[Abstract]:At present, the mainstream technology of flow recognition and classification is based on statistical method, the core of which is to extract effective feature attribute set. However, the assumption of this method is that the feature is not related and the data is not related.Because of the irrationality of this assumption, the classification effect and recognition performance are limited. The multifractal theory with data correlation as the core is introduced, and the limitation and parochialism of independent hypothesis are abandoned fundamentally, and the effective classification of flow is realized.For this reason, the fractal spectrum of the flow is defined and proved, and the estimated spectrum of the diversion flow is pushed on this basis, and then the estimated spectrum is analyzed based on the grey correlation degree in the defined kernel domain, and then the classification and recognition of the flow is realized without the feature extraction process.Finally, a series of experiments are carried out to show the fractal character and spectrum of the flow, and the vertical and horizontal comparison of the actual classification results is carried out.The research results show that the multi-fractal theory based flow classification recognition method can effectively compensate for the inevitable independent hypothesis defects of statistical methods, so it has a powerful and efficient ability to identify unknown flows.It is also especially suitable for dynamic and changeable online recognition.
【作者单位】：安徽师范大学物理与电子信息学院;南京邮电大学通信与信息工程学院;
【基金】：国家自然科学基金资助项目(61401004) 安徽省自然科学基金资助项目(1508085QF133) 安徽师范大学创新基金资助项目(901-741407)
【分类号】：TP393.06

【相似文献】