An Adaptive Ordering Framework for Filtering Multimedia Streams

In multimedia stream filtering scenario, there usually exist many filtering rules that specify the filtering objectives and many filtering units that estimate the filtering rules. A filtering rule may connect to several different filtering units and a filtering unit may connect to several different filtering rules. An open problem in such a filtering scenario is how to order the filtering units in an optimal sequence so as to decrease the filtering cost. Existing methods are based on a greedy strategy which orders the filtering units according to three factors of the filtering units, i.e., the selectivity, popularity, and cost. Although all these methods reported good results, there is still one important problem that hasn´t been addressed yet. The selectivity factor is set empirically, which is unable to adaptively adjust with stream passing by. Under these observations, in this paper, we propose an Adaptive ordering framework (AOF) which executes an adaptive ordering strategy. In AOF, all the temporal filtering results are preserved in each sliding window. Accordingly, the selectivity can adjust automatically and thus all the filtering units can be ordered with respect to the adapted selectivity. Experiments on both synthetic and real life multimedia streams demonstrate that our AOF method outperforms other simple filtering methods.