Implementation of a 2D/3D multimedia content adaptation decision engine

• Main Author: Fernandes, Rui
• Publication Date: 2015
• Other Authors: Andrade, M.T.
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10198/12584
• topic: Multimedia content adaptation decision

Multimedia content consumption has become very popular due to several factors, among which the amount of content available online and the ubiquity of network connectivity. In fact nowadays everyone can be a content producer and be almost permanently connected to the Internet, thus having the possibility to consume content anywhere, anytime. However, content may present itself in a multitude of formats and networks and terminals may offer very dissimilar transport and consumption capabilities, both along time and space. Accordingly, it often happens that the delivery and/or consumption environments do not offer sufficient or adequate resources to allow the remote access to and consumption of original high quality content. Content adaptation techniques establish means to surpass those impossibilities, allowing content delivery to the user regardless of existing constraints. Since there are several ways to adapt multimedia content, to which different users may react differently, the adaptation engine in charge of deciding the type of adaptation to perform, should ideally be driven with the aim of providing the best Quality of Experience (QoE) to the user [1]. To achieve the best possible outcome, the engine should take into consideration the characteristics of every entity and person involved in the content consumption process, which includes, (1) the multimedia content itself, (2) the transport/access network characteristics, (3) the terminal device characteristics and the (4) user preferences. To take into consideration the multimedia content it is necessary to characterise it and eventually to classify it into a set of limited but meaningful classes. Different metrics were implemented to tackle the content characteristics identification/classification. These were mainly focused on the spatial and temporal complexity classification of the content. The networks characteristics establish the restrictions the adaptation decision algorithm has to obey. This is also true for the device capabilities/characteristics. The user preferences are the subjective element that may establish, for one consumption scenario, different QoE, with the use of the same adaptations, for different users. To investigate this parameter, a subjective quality evaluation was performed. Different contents were generated and classified using the metrics devised to perform that task [2] and, based on this classification, four contents with different characteristics were chosen to be presented to the users. Several bitrates were used to simulate different network conditions, three different types of terminals were used (display, tablet and smartphone) and three adaptations were executed over the original contents, namely, spatial, temporal and quality alterations of the content. The users were asked to classify each presented version on a qualitative scale [3]. The obtained results indicate, as expected, the existence of different users profiles and that the (4) users preferences are dependent of the other three factors: (1), (2) and (3). Results from this subjective experiment are now under analysis to generate these user profiles using an approach that complements Multiple Correspondence Analysis and Cluster Identification. All these characteristics are to be used by a learning algorithm to define the cost of executing a certain adaptation, whenever a certain content is being consumed under specific conditions by a certain user. These costs are then fed to the adaptation decision engine, already implemented through a Markov Decision Process (MDP) to define the final adaptation decision. References: [1] “Definition of quality of experience,” TD109rev2 (PLEN/12), International Telecommunications Union, ITU-T Study Group 12, 2007. [2] J. Korhonen, U. Reiter, and A Ukhanova, “Frame rate versus spatial quality: Which video characteristics do matter?,” in Visual Communications and Image Processing (VCIP), 2013, Nov 2013, pp. 1–6. [3] ITU-R Recommendation BT.500-13, “Methodology for the subjective assessment of the quality of television pictures,” Tech. Rep. BT.500-13, International Telecommunications Union, January 2012.