An algebraic approach to the design of low complexity personalized video playlist

We consider the problem of automatically constructing a personalized playlist composed of a sequence of videos that attain a specific user profile. The proposed system can be used in personal multimedia devices and smartphones. Each video item is represented as a collection of weighted attributes that represent the accompanying features. The proposed solution allows users to construct a sequence of video items by imposing their preferences through Graphical User Interface (GUI). Users' preferences include inter-track relation and intra-track global features. We adopt an algebraic approach in order to model the playlist as the solution to a constrained sequence retrieval problem which leads to an integer optimization framework. Low complexity implementation is achieved by adopting the method of relaxation which leads to ordinary linear program with very low complexity. We demonstrate the flexibility of the proposed solution in constructing the playlist according to user preferences using simulation on both artificial and real video collections.