基于内容的Flash Movies分类研究

发布时间：2018-10-16 18:37

【摘要】：随着Internet的迅速发展，网络已经深入到人们生活的众多领域，成为了人们不可或缺的通讯工具、信息获取工具及娱乐工具。Flash Movies作为一种新兴的多媒体形式，是Internet中的重要组成元素，由于具有极强的艺术表现力、制作简单、交互灵活、文件小、便于网络传播等诸多优势，在Internet中广泛应用于游戏、动画、MTV、广告、教学课件，也作为文档、PPT、视频的容器出现，其数量不断增长，成为目前十分重要的多媒体网络信息资源。Flash Movies资源在教育领域所发挥的作用也越来越重要，一方面，互联网技术的发展使得这些Flash Movies资源的发布与共享不再受时间、空间的限制，成为我们获取这些资源的一个重要途径；但另一方面，网络资源的丰富且动态更新却又使我们从中寻找需要的Flash Movies资源变得越来越困难。 Flash Movies资源作为教育资源建设的一个重要方面成为教育技术学的研究领域。面对网络中越来越多的Flash Movies，如何管理才能使人们能够准确而方便的找到自己想要的资源，成为一个引人注意的问题。解决这一问题的行之有效的方法就是对其进行分类，通过分类管理、分类检索可以极大的提高管理的效率和检索的效率。因此，本文通过对Flash Movies内容特征的分析和分类算法的研究，旨在建立一个基于内容的Flash Movies分类系统，实现对Flash Movies的自动分类管理。该分类系统选取了Flash Movies的14个类别特征项来标识一个Flash Movies的类型，这些特征项分别是文件大小、变形数、图形数、文本数、声音数、按钮数、影片剪辑数、脚本数、帧数、游戏匹配度、动画匹配度、MTV匹配度、课件匹配度和广告匹配度，其中最后5项为文本特征项。这种选取方法既包含了Flash Movies的内容特征，又将文本特征纳入到分类体系之中，实现了两者的优势互补。该分类系统核心算法采用了较为成熟的BP神经网络算法，BP神经网络是一种按误差反向传播算法学习的多层前馈网络，是目前应用最广泛的神经网络模型之一。BP神经网络是对人脑的简单模仿，但它仍具有两个关键特征：第一，人工神经网络与人脑一样，都是由若干可计算单元高度连接构成的复杂网络。第二，单元之间的连接和信息处理方式决定了网络的功能。BP神经网络通过样本的训练，学习获得神经元之间连接的各项参数，主要是各连接权值和阈值，这就是神经网络的学习过程。最后将要分类的Flash Movies的各特征项的值输入到经过学习的BP神经网络中，经过计算获得其类别。本文的研究工作和研究成果主要包括以下几个方面：（1）搜索下载Internet上的Flash Movies，建立一个Flash Movies资源数据库，分析数据库中Flash Movies的类别特征元素及其属性。（2）提取Flash Movies关键词的特征向量，计算Flash Movies与每一类的相似度，将相似度纳入到Flash Movies的类别特征描述之中。（3）研究设计了适合于对Flash Movies进行分类的算法。（4）研究分析了关于BP神经网络的改进算法。（5）在理论研究的基础上，完成了一个基于内容的Flash Movies分类系统，，并将其应用于Flash Movies的资源管理平台。
[Abstract]:With the rapid development of Internet, the network has penetrated into many fields of people's life, becoming an indispensable communication tool, information acquisition tools and entertainment tools. As a new kind of multimedia form, Flash Movie is an important composing element in Internet. Because of its strong artistic expression, simple production, flexible interaction, small file, convenient network propagation and so on, it is widely used in games, animation, MTV, advertisement and teaching courseware in Internet. As a document, PPT, video container appears, its number is increasing, becomes the most important multimedia network information resource at present. The role of Flash Movie resources plays a more and more important role in the field of education. On the one hand, the development of Internet technology makes the publishing and sharing of these Flash Movie resources no longer limited by time and space, and becomes an important way for us to get these resources; but on the other hand, The rich and dynamic update of network resources makes it increasingly difficult for us to find the required Flash Movie resources. As an important aspect of the construction of educational resources, Flash Movie resources have become the research leader of educational technology In the face of more and more Flash movies in the network, how to manage can enable people to find their own resources accurately and conveniently, and become a noticeable question The effective way to solve this problem is to classify it. Through the classification, the efficiency and efficiency of management can be improved greatly by classification and retrieval. Therefore, through the research of the analysis and classification algorithm of Flash Movie content features, this paper aims to establish a content-based Flash Movie classification system, which can realize the automatic classification of Flash Movie. The classification system selects 14 category feature items of Flash Movie to identify the type of a Flash Movie, which are file size, number of deformation, number of graphics, number of text, number of sounds, number of buttons, number of movie clips, number of scripts, number of frames, match degree of game, animation match, degree, MTV matching degree, courseware matching degree and advertisement matching degree, wherein the last five items are text features The selection method not only contains the content features of Flash Movie, but also the text features into the classification system, and realizes the advantages of both. The BP neural network is a multi-layer feedforward network based on error backward propagation algorithm, which is the most widely used neural network model. One. BP neural network is a simple imitation of human brain, but it still has two key features: first, artificial neural network is the same as human brain, it is a complex network composed of several computing unit height connections. The connection between the second and second units and the information processing mode determine the work of the network Through training of samples, BP neural network learns the parameters of connection between neurons, which is mainly the value of each connection and the threshold, which is the learning of the neural network. and finally, inputting the values of each feature item of the Flash Moies to be classified into the learned BP neural network, Don't. The research work and research results of this paper mainly include the following: Aspect: (1) Search the Flash Movie on the Internet, set up a Flash Movie resource database, analyze the category feature elements of Flash Movie in the database and (2) extracting feature vectors of the Flash Movie keywords, calculating the similarity between the Flash Movie and each class, and integrating the similarity into the classes of the Flash Movie. Do not characterize. (3) The study is designed to be suitable for Flash Movi Algorithms for classifying es. (4) Study and analysis about The improved algorithm of BP neural network. (5) On the basis of theoretical research, a content-based Flash Movie classification system is completed and applied to Flash.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP37

【相似文献】