With the emergence of advanced video encoding technology and the rapid deployment of broadband wireless networks, high quality wireless video services become increasingly feasible. Transporting of video traffic over the broadband wireless networks is likely to be a dynamic industry. A critical issue for multimedia applications on wireless is to ensure that the quality-of-service (QoS) is maintained at an acceptable level. Furthermore, this QoS must be ensured under unreliable and time-varying wireless channels. Thus it is particularly important to develop analytical performance model for wireless video transmission systems....[ Read more ]

With the emergence of advanced video encoding technology and the rapid deployment of broadband wireless networks, high quality wireless video services become increasingly feasible. Transporting of video traffic over the broadband wireless networks is likely to be a dynamic industry. A critical issue for multimedia applications on wireless is to ensure that the quality-of-service (QoS) is maintained at an acceptable level. Furthermore, this QoS must be ensured under unreliable and time-varying wireless channels. Thus it is particularly important to develop analytical performance model for wireless video transmission systems.

In this research we examine performance issues related to wireless video communications. The study focuses on the link level performance of video transmission over the uplink of an unreliable wireless channel. The work consists of five parts. In the first part of the work, we formally define the discrete time batch Markovian arrival process (DBMAP) with marked transitions, which is featured with the versatility of a normal DBMAP in modeling correlated and bursty data traffic, and the flexibility of the Markov process with marked transitions in modeling more than one class of data arrivals. In the second part of the work, we show that the marked DBMAP can be used to model video traffic such as MPEG-4, by taking into account the inherent nature of the adaptiveness of the video traffic. We propose a priority based scheduling protocol with automatic repeat request (ARQ) control for video data transmission, by taking advantage of the adaptability of MPEG-4 video traffic. We prove that in a hidden Markov modeled (HMM) wireless channel with probabilistic transmission, the service time for an arbitrary radio link control (RLC) data burst follows discrete time phase (PH) type distribution. Extensive simulations are carried out to study the queueing behavior of the video data transmission buffer, as well as the error and drop behavior at both the RLC burst and video packet level. In the third part of the work, we formulate the DBMAP/PH/1 priority queue, in which the arrival process is a DBMAP with two priority levels, and the service time obeys the discrete PH distribution. We derive the transition probability matrix of the underlying Markov chain and show that the Markov chain is M/G/1 type. Based on matrix analytic methods, we develop computation algorithms for the probability distribution of the system numbers, and derive the key performance indices of the DBMAP/PH/1 priority queueing model. In the forth part of the work, we study video traces with layered encoding and propose a traffic model for scalable video encoded in multiple layers. The model is based on the Markovian arrival process with marked transitions. The state of the Markovian arrival process is derived from the correlation feature found in the video data. The base layer and enhancement layer video frame size pairs are classified by a cluster detection algorithm. Each cluster corresponds to one state of the underlying Markov chain of the video traffic arrival process. The joint base and enhancement layer video frame size distribution for each state of the Markov chain is approximated by multivariate normal distribution. The accuracy of the source model is validated by simulation study. In the fifth and final part of the work, we develop algorithms to estimate how much bandwidth is needed for layered video transmission over an HMM wireless channel. We show that the video data transmission buffer can be modeled as a G/D/c queue with time dependent feedback. We propose to approximate the feedback traffic by an HMM modulated DBMAP. Based on the effective bandwidth approach, we derive the required channel capacity in order to meet the given QoS targets for the video traffic. Numerical results for the bandwidth estimation algorithms are presented.