Variable Bit Rate Video Sources Research Articles

Modelling H.264/AVC VBR video traffic: comparison of a Markov and a self-similar source model

We compare the adequacy of two models for realistic video sources, namely the fractional Brownian motion (fBm) and Markov modulated fluid flow (Mmff) models. We use the effective bandwidth approach to get the probability that the buffer content exceeds a certain threshold. We use a formula in which the variance function, i.e., the variance of the traffic arriving in an interval of length ?, plays a central role and we model it as variance associated either with the Mmff or fBm model. We measure the variance function for an artificial source we construct, using Variable Bit Rate (VBR) H.264/AVC (Advanced Video Codec) video traces of real movies. There is a good correspondence between the buffer threshold exceeding probability obtained via trace-based simulations and the one predicted theoretically, based on the measured variance function. These we take as benchmark results against which we check both models--fBm and Mmff. First, we try to tune the model parameters such that their variance function matches the measured one over a large range of ? values, but this proves to be difficult. When matching the variance function over a short range of the ? of interest, we conclude that the Mmff model is better suited to model VBR video sources. We conduct an error sensitivity analysis with non-optimal model parameters and conclude that they do not influence considerably the buffer exceeding probability. Most reliable results are achieved with the measured variance function; if by some reason it needs to be modelled, the Mmff is preferred over the fBm model.

Neuro-Fuzzy Control for MPEG Video Transmission Over Bluetooth

The application of a neuro-fuzzy (NF) controller to moving picture expert group (MPEG-2) video transmission over a Bluetooth asynchronous connectionless (ACL) is presented in this paper. MPEG variable bit rate (VBR) data sources experience unpredictability, long delay, and excessive loss, due to sudden variations in bit rate. Therefore, it is practically impossible to transmit MPEG-2 VBR video sources over a Bluetooth channel with a limited transmission speed and considerable wireless interferences without loss of data or image quality degradation. In this work, a traffic-shaping buffer is introduced before the host controller interface (HCI) of the Bluetooth protocol stack to obstruct excessive MPEG VBR data sources during the peak transmission periods. An integrated neuro-fuzzy controller monitors and reduces the burstiness of the traffic-shaper output rate to facilitate the MPEG VBR video sources to conform to the token-bucket contract before entering the Bluetooth channel. A conventional rule-based-fuzzy (RBF) controller oversees the arrival rate to the traffic-shaper to avoid saturation or starvation of the buffer. The computer simulation results demonstrate that applications of the NF and RBF schemes reduce excessive delay and data loss at the HCI as compared with conventional video transmission in Bluetooth ACL links

Effective bandwidth estimation and testing for Markov sources

This work addresses the resource sharing problem in broadband communication networks that can guarantee some quality of service (QoS), and develops some results about data source and traffic modelling, especially in aspects of model testing and parameter estimation. The multiplexing of variable bit rate (VBR) sources poses a mathematical and statistical problem: the estimation of the resource requirements of a source or set of sources. The estimation method shall be simple enough to be practically implemented in the connection acceptance control (CAC) function. In this paper, the VBR video sources are taken as a typical case of variable rate, with real-time constraints. This association of requirements makes the case especially interesting. A Markov model is assumed for the VBR sources. The validity of such models is under research; they seem to be appropriate at least in certain time scales. The model is tested against real video traces. In order to estimate the resource allocation or “channel occupation” of each source, the concept of equivalent bandwidth proposed by Kelly [Notes on effective bandwidth, in: F.P. Kelly, S. Zachary, I.B. Ziedins (Eds.), Stochastic Networks: Theory and Applications, Oxford University Press, Oxford, 1996, pp. 141] is used; it is based on a consistent mathematical theory, and has proven to be robust and useful for technical applications. A calculation of the equivalent bandwidth of a Markov source, given its parameters, can be found in the literature [IEEE ACM Trans. Networking 1 (4) (1993) 424]. But in fact, one can only estimate model and parameters. In this work, an estimation of the equivalent bandwidth is given, which can be obtained from real data. The convergence and the consistency of the estimation are studied, and practical bounds are found. Illustrative calculations are performed from real video traces that were obtained using a software MPEG coder, developed by the authors. The mathematical and statistical results are valid for whatever phenomenon that can be modelled as a Markov process.

A general AR-based technique for the generation of arbitrary gamma VBR video traffic in ATM networks

Modeling variable-bit-rate (VBR) video source traffic is a crucial issue to evaluate the end-to-end performance of transmitting video signals over asynchronous transfer mode (ATM) networks. Difficulties in source modeling arise from the fact that VBR video source traffic usually follows a gamma distribution with high correlation among adjacent data. Many researchers adopt autoregressive (AR) models driven by a Gaussian error process to account for such correlation. The problem is: due to the closure property of the Gaussian distribution, the traffic so generated is Gaussian rather than gamma. As a remedy, some researchers directly consider gamma AR models instead. Unfortunately, the trouble arises from the fact that the closure property does not apply to gamma distributions, and thus, the linear operation performed by an AR model fails to produce gamma traffic. In this paper, we present a new technique that is capable of generating gamma-distributed traffic with arbitrary correlation while retaining the computational efficiency of Gaussian AR models. The central idea is to decompose given gamma traffic into a linear combination of a number of /spl lambda//sup 2/(1) sequences, and each of these latter processes can be easily obtained from a Gaussian AR process through a simple nonlinear operation. Results based on actual video teleconference data are presented to verify the validity of the new algorithm.

The importance of long-range dependence of VBR video traffic in ATM traffic engineering

There has been a growing concern about the potential impact of long-term correlations (second-order statistic) in variable-bit-rate (VBR) video traffic on ATM buffer dimensioning. Previous studies have shown that video traffic exhibits long-range dependence (LRD) (Hurst parameter large than 0.5). We investigate the practical implications of LRD in the context of realistic ATM traffic engineering by studying ATM multiplexers of VBR video sources over a range of desirable cell loss rates and buffer sizes (maximum delays). Using results based on large deviations theory, we introduce the notion of Critical Time Scale (CTS). For a given buffer size, link capacity, and the marginal distribution of frame size, the CTS of a VBR video source is defined as the number of frame correlations that contribute to the cell loss rate. In other words, second-order behavior at the time scale beyond the CTS does not significantly affect the network performance. We show that whether the video source model is Markov or has LRD, its CTS is finite, attains a small value for small buffer, and is a non-decreasing function of buffer size. Numerical results show that (i) even in the presence of LRD , long-term correlations do not have significant impact on the cell loss rate; and (ii) short-term correlations have dominant effect on cell loss rate, and therefore, well-designed Markov traffic models are effective for predicting Quality of Service (QOS) of LRD VBR video traffic. Therefore, we conclude that it is unnecessary to capture the long-term correlations of a real-time VBR video source under realistic ATM buffer dimensioning scenarios as far as the cell loss rates and maximum buffer delays are concerned.

Admission control for variable bit rate video in Banyan switches

When a Banyan switch is used to switch variable bit rate (VBR) video, some video packets may not be switched before their deadlines because the VBR video sources may temporarily generate heavy traffic or they only generate moderate traffic but some video packets encounter internal blocking in the switch. Therefore, some video packets may be discarded. We determine the condition of packet loss and then derive the packet loss probability. Based on this result, we perform admission control to determine whether a given video session can be admitted while the resulting packet loss probability is still small enough to guarantee a good image quality.

Study of the impact of MPEG-1 correlations on video-sources statistical multiplexing

Variable bit-rate (VBR) video is expected to become increasingly important with the large scale deployment of broadband-integrated services networks (B-ISDNs). While the modeling of VBR video sources has received significant attention, there is no widely accepted model which lends itself to mathematical analysis. Furthermore, new video compression standards are emerging, such as the MPEG family. A detailed statistical analysis of a long sample of a movie encoded with an MPEG-1 algorithm highlights that traffic generated by an MPEG-1 encoder has both a high degree of burstiness (peak/average ratio greater than five) and a strong long-term correlation (positive correlation lasts for ten/twenty minutes). In this paper, we propose and analyze a Markov model which captures both the burstiness and the temporal dependencies of MPEG traffic. Furthermore, we show that our model is a flexible traffic generator which is suitable for studying statistical multiplexing issues. Specifically, we study both the multiplexing of a VBR video source with low priority traffic and the multiplexing of several VBR video sources. The results obtained show that the long-term correlation significantly affects the tail of the delay distribution even for lightly loaded networks. The presence of a long range dependence in the MPEG traffic significantly reduces the gain in terms of network utilization which can be achieved with statistical multiplexing of i.i.d. sources.

Modeling activity in VBR video sources

Variable bit-rate (VBR) video sources are significantly characterized by the statistics of ‘scene changes’, which determine the activity exhibited by the pictures. In a source coded with an interframe technique, in fact, this characteristic is related to the length of the emission periods having a very high bit-rate. It therefore represents a fundamental feature in evaluating performance of a network supporting video traffic. In this paper an emission process that is able to fit different activity measurements for VBR video sources characterized by a predefined bit-rate probability density function is introduced. Such a process is based on a suitable linear combination of continuous state autoregressive Markov processes. It is analyzed with respect to its capability to match the actual values of some temporal parameters which have been introduced to characterize the source. Its modeling power is compared with that of a single autoregressive Markov process. In particular is demonstrated that the process introduced here can model faster motion within a scene than a single autoregressive Markov process. Finally a case study is introduced to point out the flexibility of the proposed process.