A personalized system for scalable distribution of multimedia content in multicast wireless networks

Abstract

This paper presents a novel architecture for scalable multimedia content delivery over wireless networks. The architecture takes into account both the user preferences and context in order to provide personalized contents to each user. In this way, third-party applications filter the most appropriate contents for each client in each situation. One of the key characteristics of the proposal is the scalability, which is provided, apart from the use of filtering techniques, through the transmission in multicast networks. In this sense, content delivery is carried out by means of the FLUTE (File Delivery over Unidirectional Transport) protocol, which provides reliability in unidirectional environments through different mechanisms such as AL-FEC (Application Layer Forward Error Correction) codes, used in this paper. Another key characteristic is the context-awareness and personalization of content delivery, which is provided by means of context information, user profiles, and adaptation. The system proposed is validated through several empirical studies. Specifically, the paper presents evaluations of two types that collect objective and subjective measures. The first evaluate the efficiency of the transmission protocol, analyzing how the use of appropriate transmission parameters reduces the download time (and thus increasing the Quality of Experience), which can be minimized by using caching techniques. On the other hand, the subjective measures present a study about the user experience after testing the application and analyze the accuracy of the filtering process/strategy. Results show that using AL-FEC mechanisms produces download times until four times lower than when no protection is used. Also, results prove that there is a code rate that minimizes the download time depending on the losses and that, in general, code rates 0.7 and 0.9 provide good download times for a wide range of losses. On the other hand, subjective measures indicate a high user satisfaction (more than 80 %) and a relevant degree of accuracy of the content adaption.