小波域基于分段Hurst指数的视频流分类

发布时间：2018-04-21 05:54

本文选题：视频流 + 分类　；参考：《电子与信息学报》2017年06期

【摘要】：现有的视频流分类方法体现出内容依赖及特征依赖的局限性,该文引入流量分形理论,并在小波域内,提出一种基于Hurst指数的Fractals分类模型以改进不足。为此,该文首先描述流的分形性质,定义流的Hurst指数,推导小波域内Hurst指数的估计过程。然后,基于代价函数优化分段目标,用聚类差异度方法计算分段Hurst指数的总体差异量,再基于最大类间方差阈值进行分析,从而实现视频流的细粒度分类。研究结果表明,该文提出的分类方法,以随机数据的变化特性为内容,突破了内容依赖的局限性,解决了特征制约的瓶颈,提高了视频流的分类效果。
[Abstract]:The existing video stream classification methods reflect the limitations of content dependence and feature dependence. In this paper, the flow fractal theory is introduced, and a Fractals classification model based on Hurst exponent is proposed in wavelet domain to improve the model. This paper first describes the fractal properties of the flow, defines the Hurst exponent of the flow, and deduces the estimation process of the Hurst exponent in the wavelet domain. Then, based on the cost function to optimize the segmented target, the total difference of segmented Hurst exponent is calculated by clustering difference degree method, and then based on the maximum inter-class variance threshold, the fine-grained classification of video stream is realized. The results show that the proposed classification method, which takes the variation of random data as the content content, breaks through the limitation of content dependence, solves the bottleneck of feature restriction and improves the classification effect of video stream.
【作者单位】：南京邮电大学通信与信息工程学院;安徽师范大学物理与电子信息学院;
【基金】：国家自然科学基金(61271233,60972038,61401004) 华为HIRP创新项目~~
【分类号】：TN919.8

【相似文献】