This paper introduces a novel content classification scheme in video-on-demand (VoD) based on the emerging standard of MPEG media transport (MMT). Content classification is necessary and reasonable, due to the fact that a better user experience may be obtained if the requested video is organized and presented in different ways according to various user requirements and/or video popularities. For this purpose, it is proposed that under the MMT framework, the requested video needs to be partitioned into chunks, which are labeled by level ID at the sender, and rearranged by the composition information (CI) in MMT at the user terminal. With this scheme, VoD service provider can easily offer multiple versions of the same video, which are convenient for users to choose from on the basis of individual preferences.
The overlapped content in multiple video versions do not have to be re-delivered if requested. The server just needs to send the absent chunks of the requested version, and integrate them with the already received and locally cached chunks under the guidance of CI. Significant traffic reduction and bandwidth saving can be achieved. We may also expect decreased buffering time at the user terminal and reduced storage usage at the server. The proposed scheme has been implemented over an experimental MMT demonstration system, and its validity has been verified.