Therefore, there is a critical need for proper data models to organize and manage massive multimedia data, which should achieve to provide multimedia information retrieval service for digital libraries by means of multi-level semantic data mining. According to such needs, this paper proposes an efficient data model for multimedia content retrieval. Section 2 presents the detail of our data model.
2. Related Work 2.1. The Multimedia Data Modeling Compared to the ordinary multimedia data structured data is much more complicated, first the structure of multimedia data is semi-structured data (such as XML and HTML) and unstructured (text, images, audio and video), in addition the large amounts of multimedia data content information it carries, people with different background knowledge has various understanding of the content [6]. The purpose of the organization and structured multimedia database system is for content retrieval [7]. In this sense, the construction of multimedia data model is one of the key factors. The method for modeling multimedia data, it usually can be divided into the following three kinds of modeling methods: extended relational model modeling method, the semantic model modeling method and object-oriented modeling method [8].In recent years, aimed to the multimedia database system and multimedia management has appeared a lot of data model. In [9] the algebra video (algebraic video) data model is one of the earliest and fairly complete data model, which has a great influence in the following models.This model can distinguish between physical streaming and logic video segments, and also allow users to synthesis of parallel video performance, which is provided by the support of video algebra (video algebra). However, this model is very difficult to join automatic video segmentation methods. H. Jiang proposed a video Data Model called Logical Hypervideo Data Model(LHVDM) [10].This model is able to perform multiple levels video extraction (present the video object of interest to the users), it is referred to as hot objects, and also it put the extracted semantic and other logic extraction together(including other hot objects). DISIMA [11] is the image database system developed by the university of Alberta, which proposes a model applicable to image and space applications. The generic indexing model - VIDEX is proposed in [12], that is applied to the multimedia information system SMOOTH [13].This method combines the advantages introduced by [9-10], introduces the video streams structured approach and indexing process versatility extends in these models. Farham Mohamed [14] proposed a data model based on temporal to manage the multimedia data. Besides Liu presented a multi-media data organization and management model which based on object-oriented [15]. Thought the current multimedia data model solved the related problems about modeling a single media, many different media integrated modeling, the correlation analysis between different media data and seamless integration in digital library still did not perform well. Object-oriented method provides an expression, management, and share options for multimedia data modeling scheme. However, a monomedia oriented object model is very difficult to meet the needs of the digital library, which is in the demand of a variety of media data organization and management. Digital library needs to deal with a variety of types media data, and these different types of media data for unified modeling and management, therefore need a more powerful multimedia data model to support effective modeling and seamless integration with digital library application demand, realize all levels semantic mining of the multimedia data and provide efficient digital book data retrieval service.