Hidden Markov models for packet-level errors in bursty digital wireless channels

Packet-level generative models are designed to be employed for evaluating and designing high layer wireless communications protocols. The significance of these models arises from their ability to replace the entire physical communication layer, hence reducing the time and complexity of wireless system simulations. In this paper, an enhanced general packet radio service (EGPRS) transmission system is utilized to obtain reference bursty packet error sequences. We then propose a new hidden Markov model with two processes to generate packet error sequences of any desired length. The attained burst error statistics demonstrate a satisfactory performance compared to those of reference packet error sequences.