中文摘要：

针对目前媒资管理系统依赖编目信息进行检索而出现的编目信息难以覆盖媒资数据的所有语义内容、由于人的理解不同而导致的编目信息不一致、媒资编目费力费时等问题，设计了不依赖编目信息的基于全文检索、语音识别、人脸识别、关键帧提取等的智能媒资检索系统，对媒资内容自动分析、媒资特征索引、媒资特征检索进行了阐述，并采用基于B／S的分布式架构进行了实现。结果证明，该方案设计具有较高的可靠性和稳定性，在媒资管理中得到了良好的应用。

英文摘要：

In view of the problems of the catalogue based retrieval in current media asset management systems： 1 ） catalogues cannot cover all the semantic information contained in the media data; 2 ） inconsistance of catalogues due to different understandings of different people; 3） time-consuming and inconvenience of cataloguing, a catalogue-independent intelligent media asset retrieval system is designed based on full-text search, speech recognition, face recognition, key-frame detection in this paper. And then, technical details of media content analysis, media feature indexing and searching are described. At last, a distribution system is established based on B/S architecture to verify our proposed methods. The experimental results prove that the design and implementation are efficient and effective, and the system has good applications in media asset management.