Abstract: The paper constructs a model for real-time raw multimedia data mining and proposed a new mechanism.We apply multi-level vector distance to process and control sequence analysis segmentation and feature extraction,and uses the novel K-means clustering based on Particle Swarm Optimization algorithm。With proposed technologies, this model can solve some special demand of multimedia data mining at a cetain extent.

摘要： 本文构造了一个实时多媒体数据挖掘模型，提出了一种原始视频数据进行数据挖掘的新机制，主要采用分层向量距离来进行动态可控序列分析分割、段内特征提取，使用粒子群K均值进行段间聚集，在一定程度上，解决了多媒体数据挖掘各方面的特殊要求。挖掘模型中各个部分与本文提出的技术相结合，基本上能满足实时情况下处理原始视频数据的要求。