Combined Markov Model for Packet Loss Characterization in UMTS Channels

Mobile communication channels such as UMTS channels are affected by errors which show certain statistical dependencies. In this paper we tackle the characterization of error occurrences from a packet level perspective. We present a novel approach which consists of a generative model based on the interconnection of a Markov gap model, aimed to characterize error-free bursts, with several hidden Markov models (HMMs), which characterize the distribution of bursts with erroneous packets. We demonstrate that this model provides accurate predictions on UMTS radio links and compare its accuracy with that of traditional models based on Markov chains.