Proper

Abstract

Mobile devices are susceptible to changing bandwidth when moving. Current HTTP-based adaptive streaming deals with such bandwidth situation by providing an appropriate video representation for current available bandwidth. Recent studies on adaptive streaming showed that solving network problems with predictions could improve the quality of service. Therefore, in this paper, we propose a method of predicting downloading segments to solve the problems of frequent network changes. In this method, we first calculate the average of previous bandwidth values, the average time to download previous segments. Second, we trace-back at some earlier bandwidth values, average downloading time values to infer a degree of current network fluctuation. We then predict a new request segment for a mobile client by using fuzzy logic. In the experiment, our results show that the method significantly reduces the fluctuation compared to previous research.