DEVELOPING AUDIO PROCESSING AGENTS FOR MULTI-AGENT MPEG-7 ENABLED ENVIRONMENT Mingkun Li + , Gang Wei * , Valery A. Petrushin * , Ishwar K. Sethi + + Department of Computer Science and Engineering Oakland University Rochester, MI 48309, USA {li,isethi}@oakland.edu * Accenture Technology Labs 161 N. Clark Street, Chicago, IL 60601, USA {gang.wei,valery.a.petrushin}@accenture.com ABSTRACT This paper presents a methodology for developing audio processing agents for a multi-agent environment that is known as the Community of Multimedia Agents. The Community’s philosophy, objectives and architecture are described. The methodology is illustrated using audio feature extraction agents as example. The algorithms used for extracting audio features are classical and work in general audio domain. The agents have the standard MPEG-7 interface for better interoperation and wide usage. Two low level tools - the MPEG-7 audio descriptor wrapper classes and the MPEG audio decoder – are also presented. An example of agent aggregation for an annotation system prototyping is provided. 1. INTRODUCTION Multimedia data is growing explosively and the techniques to find the desired content is lagged far behind despite of the amount of researcher efforts. Challenges in video, audio and image content annotation are making the collaboration between researchers increasingly important, which is however held back by a number of factors, such as the variety of platforms, programming languages, and data exchange formats and unwillingness of corporations to distribute their intellectual property unprotected. For example, when one wants to build a multimedia analysis system that requires face detection, very likely he or she will end up re-implementing some existing algorithm or creating his/her own, even though dozens of them have already been developed. On the other hand, when one invents a new tool, the chances of it to be used by others depend heavily on its exposure to web search engines and other indexing tools. Academic researchers would usually be happy to get ready their tools for sharing if it does not take too much effort. This situation triggered the Community of Multimedia Agents project at the Accenture Technology Labs [1]. 2. THE COMMUNITY OF MULTIMEDIA AGENTS The Community of Multimedia Agents is a virtual community of researchers and students, who exchange their multimedia analysis tools and test data. The Community’s objective is to improve the cross- organization collaboration, consolidate efforts and expedite research and education in the field of multimedia analysis and annotation. The Community’s organization is based on the following principles. • Web-based Portal. The Community is located on the World Wide Web at http://community.techlabs.accenture.com and accessible from any Internet-able workstation. Beside the tools (see below) the Community Portal provides information about related conferences and workshops, business and academic news, links to related projects, book and paper recommendations, tutorials, etc. • Free Membership. The Community’s membership is free. • Media Library. The Community provides links to media files that could be used for multimedia annotation research. The Community members provide these links to the collections of media data. The copyright belongs to the authors of the collections or/and their organizations