数字化音乐资源情感检索技术研究

发布时间：2018-02-21 09:14

本文关键词： 音乐元数据 AHP法情感检索模糊数学　出处：《华中师范大学》2013年硕士论文　论文类型：学位论文

【摘要】：多媒体技术和计算机技术的交互促使了数字化音乐资源在人们的生活中被频繁应用。娱乐、学习、广告等多个领域也提出了更加丰富的音乐视听需求。如何有效组织和查询数字化音乐资源成了亟待解决的问题。在信息资源的组织和查询中,元数据标准是基础,查询方法的质量是关键。虽然业内已经建立了诸多数字资源描述的元数据标准,但对描述数字化音乐资源的元数据还没有进行规范化定义。目前,除了几种常见的音乐检索标签,情感检索也逐渐成为人们休闲娱乐、工作学习范围内的惯用的资源获取方式。但音乐情感具有不确定性和模糊性,传统的检索方法已经难以满足人们多样化的需求,也不能凸显音乐情感的特性,故寻找新的音乐情感检索方式也成为急需解决的问题。针对上述两个问题,本文首先用文献法分析了数字化音乐资源的定义、范围、特点,并在参考DC、LOM等元数据方案的基础上,提出了具有3个层次的数字化音乐资源元数据规范。其次,根据元数据规范的设计流程从多方面进行元数据置标,如定义、版本、数据类型等。并结合RDF图和XML/RDF代码用具体范例说明如何运用该方案。成功建立元数据标准之后,又以音乐的情感特性和模糊数学原理为切入点。借助国外专家提出的音乐情感空间模型——AV模型,设计出了四种基本类情感的隶属度函数。并借助层次分析法AHP,凝聚多个音乐专家的经验和智慧,探讨了活跃度A和诱发力V在决定音乐情感隶属度值时的权重,力求使隶属度函数更加贴近实际。最后,搭建了实验平台,选取100个实验样本,从查全率,查准率等多个角度验证了该检索方法的有效性。这两个问题的解决,不仅为推动资源的共享和传播速度打下了良好的基础,也是促进资源应用深度和广度的基建工作,能为数字化音乐资源的后续应用打下了坚实的基础。
[Abstract]:The interaction of multimedia technology and computer technology promotes the digital music resources to be used frequently in people's life. Advertising and other fields also put forward more abundant music audiovisual requirements. How to effectively organize and query digital music resources has become a problem to be solved. In the organization and query of information resources, metadata standards are the basis. The quality of query method is the key. Although many metadata standards for digital resource description have been established in the industry, the metadata describing digital music resources has not been standardized. In addition to several common music retrieval labels, emotional retrieval has gradually become a popular way for people to obtain resources in the scope of leisure, work and study. But music emotion is uncertain and fuzzy. Traditional retrieval methods have been difficult to meet the diverse needs of people, and can not highlight the characteristics of music emotion, so finding a new way of music emotion retrieval has become an urgent problem. In order to solve the above two problems, this paper first analyzes the definition, scope and characteristics of digital music resources by the method of literature, and puts forward the metadata specification of digital music resources with three levels on the basis of referring to the metadata schemes such as DCLOM. Secondly, according to the design flow of metadata specification, metadata is set up from many aspects, such as definition, version, data type and so on, and the RDF diagram and XML/RDF code are used to illustrate how to use the scheme. After the successful establishment of the metadata standard, the emotional characteristics of music and the principles of fuzzy mathematics are taken as the breakthrough point. With the help of the music emotion space model proposed by foreign experts, the AV model, The membership function of four basic kinds of emotion is designed. With the help of AHP AHPand the experience and wisdom of many music experts, the weights of activity A and evoked force V in determining the membership value of music emotion are discussed. Strive to make the membership function closer to the reality. Finally, an experimental platform is built, and 100 experimental samples are selected to verify the effectiveness of the retrieval method from several angles, such as recall rate and precision rate. The solutions to these two problems not only lay a good foundation for promoting the sharing and spreading speed of resources, but also promote the capital construction work of the depth and breadth of resource application, and lay a solid foundation for the subsequent application of digital music resources.
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3

【参考文献】