基于Web Service的多层音频分类器设计

Design of multi-tires audio classifier based on web service

随着网络技术的发展和多媒体数据的增长,应用对音频内容的分类、查询、检索需求越来越强烈。MPEG-7和MPEG-21等标准的出现为多媒体内容的规范化提供了良好的外部条件,有利于进一步向基于内容检索服务的过渡。介绍了一个基于Web服务分布式部署的多层音频分类器的设计,提出以特征代理为前端进行特征提取,分类代理为后端进行特征处理,节点采用多分类器融合策略,通过发布/订阅机制实现音频特征的传送。系统结构有利于进一步扩展,从而为音频内容查询检索和音频搜索引擎等应用提供平台式服务。

With the development of web technology and the increase of multimedia data, there has been huge demand for easily manipulating on audio data via network including classifying, searching and indexing. It＇s naturally for bethinking of building a content-based audio search engine. Undoubtedly, the issue of MPEG-7 and MPEG-21 will promote this progress. An audio content classifier platform based on web service was introduced, in which a kind of feature extractor agents being front and classifier agents being back. A scheme of publishing and subscribing was also developed for audio content transmission. The architecture could be flexibly expanded for providing services for audio data search engine in future.