Introduction

The research efforts in the ICT Group in the field of multimedia content management address the following challenges:

• automating multimedia content indexing and retrieval processes,
• enabling quick, easy and personalized access to multimedia content.

To meet the first challenge, we combine our expertise in multimedia signal processing and machine intelligence with state-of-the-art achievements in the fields of "traditional" information retrieval and human perception and aim at bridging the gap between the measurable properties (features) of multimedia signals and the content conveyed by these signals. The second challenge is pursued, on the one hand, by optimizing the way multimedia content is stored, organized, abstracted and represented and, on the other hand, by developing methods for reliably learning user preferences and for filtering, pruning, adapting and delivering multimedia content accordingly.

Example

With the advent of digital video revolution the television broadcasting industry is slowly but surely transferring to an end-to-end digital television production, transmission and delivery chain. Supported by the availability of broadband communication channels, this transfer will lead to an enormous increase in the amount of video data reaching our homes. At the same time the quickly growing capacity-versus-price ratio of digital storage devices is likely to make such devices highly popular with consumers. A combination of the abovementioned phenomena will result in an explosion in the “consumer choice”, that is, in the number of video hours that are instantaneously accessible to the consumer. This may have crucial consequences for the ways the broadcasted material is “consumed”. For instance, the understanding of the broadcasting mechanism may change. This mechanism will only be something that provides data to the - soon inevitable - home mass storage system (HMSS), and as far as the consumer is concerned, the concept of the “broadcasting channel” will lose its meaning. Further, due to the large amounts of incoming data, video recording will be performed routinely and automatically, and programs will be accessed on-demand from the local storage. Viewing of live TV is therefore likely to drastically diminish with the time.

The challenge of securing the maximum transparency of the recorded video volume toward the consumer - independent of the volume size – could be approached by developing video-recommender functionality of a home mass storage system. As indicated in the figure above, this functionality would typically contain the following two main algorithmic modules:

• Module for automatically abstracting video
• Module for matching the incoming video material with user preferences.

The purpose of an algorithm for video abstraction in the context of a video-recommender functionality can be twofold. First, the abstraction algorithm can be designed to summarize the broadcasted material in order to facilitate the consumer’s choice of what to watch later on. This may be highly valuable, for instance, in the process of digesting a large volume of news television broadcasts and presenting to the user in a compact but comprehensive way the coverage of all news topics found in the volume. Alternatively, a video abstraction algorithm can be designed to prune the recorded video material by keeping the most interesting segments – highlights - only, and by discarding the remaining, less interesting parts. For instance, pruning is particularly applicable to sport broadcasts as the events being worth watching (e.g. goals in soccer, home runs in baseball, touchdowns in football) are sparse and spread over a long period of time. The second module, also referred to as the personalized video delivery module, addresses the problem of storing and organizing the incoming video material according to the subjective preferences of the consumer. Ideally, these preferences are stored in the user profile that is acquired in a non-invasive fashion, that is, without requesting complicated or uncomfortable actions from the consumer.