B. Melamed and P. Jelenkovic, "Automated TES Modeling of Compressed Video," Proc. IEEE INFOCOM '95,  Home/Search   Document Not in Database   Summary   Related Articles   Check

....can generate processes with different correlation structures with uniform probability distributions. TES models use a heuristic search to find the best TES process that can capture both auto correlation structure and distribution of the empirical data. Recently, an automated solution was presented [39]. Examples of long range dependent traffic models are given in (the fourth section) These include fractional Brownian motion, aggregate of ON OFF high variability sources, and FARIMA. Fractional Brownian motion has only one parameter controlling the auto correlation function. Therefore, there is ....

B. Melamed and P. Jelenkovic, "Automated TES Modeling of Compressed Video," Proc. IEEE INFOCOM '95,

....of long range dependence and burstiness. There is an on going debate as to how relevant LRD is to network performance when resources are limited [22,23,25,52] VBR video can be modeled using a variety of models (short range and long range) including such exotic schemes as random maps on a circle [27]. What is commonly accepted in the midst of this diversity is that compressed video is bursty, much more so than Poisson trac. 20 6.2 Performance Results 6.2.1 Unimodal Redundancy recovery Relation Figure 8 (left) shows the measured redundancy recovery function for the Simpsons MPEG I clip ....

P. Jelenkovic and B. Melamed. Automated TES modeling of compressed video. In Proc. IEEE INFOCOM '95, pages 746-752, 1995.

.... voice activity factor is between 40 and 60 of the connection time [DeV93] The duration of a phone call is assumed as exponentially distributed with a mean of 3 minutes ( Zel94] DJ91] To model video streams we used a method referred to as TES (Transform Expand Sample) modelling ( RMR94] [JM95]) We took video sequences with two different characteristics, TV broadcast like video and a video conferencing scenario, i.e. video sequences with small changes in the pictures background. These were encoded with H.261 H.263 5 and MPEG 6 1 codecs with two resolutions, QCIF 7 (176 x 144 ....

P. R. Jelenkovi'c and B. Melamed. Automated TES Modeling of Compressed Video. In Proceedings of IEEE INFOCOM95, April 1995.

....tweaks the distribution until the autocorrelations of the input process match the desired autocorrelations. Experience is required to adjust the distribution in a systematic way. TEStool is described in Melamed, Hill and Goldsman [9] An extenstion to TEStool, described in Jelenkovic and Melamed [4], provides automated fitting with respect to a particular family of distributions for V t when F Y is restricted to be the empirical cdf. Regrettably, the TEStool software is no longer available. Comparing TES to ARTA, ARTA processes are guaranteed to match p 1 autocorrelations automatically ....

Jelenkovic, P. R., and B. Melamed. 1995. Automated TES Modeling of Compressed Video. In Proceedings of the IEEE INFOCOM 1995 Conference, 746--752. IEEE Computer Society, Washington, D.C.

....dependence and burstiness. There is an on going debate as to how relevant LRD is to network performance when resources are limited [23, 25, 27, 57, 49] VBR video can be modeled using a variety of models (short range and long range) including such exotic schemes as random maps on a circle [31]. What is commonly accepted in the midst of this diversity and controversy is that compressed video is bursty, certainly much more so that Poisson traffic, and it is only this aspect that will be reflected here. 5.3 Performance Measurements 5.3.1 Unimodal Redundancy Recovery Relation Figure 19 ....

P. Jelenkovic and B. Melamed. Automated TES modeling of compressed video. In Proc. IEEE INFOCOM '95, pages 746--752, 1995.

.... These works have shown that long range dependence is an ubiquitous phenomenon occurring in diverse network environments, and even individual traffic sources such as variable bitrate (VBR) video have been shown to possess correlation structure that can be modeled along with several other models [11, 15, 13, 19, 34] using self similar traffic models. This also represents an instance of a real world traffic source which, on its own, exhibits self similar burstiness as a source intrinsic property. In the second category are papers that have evaluated the effect of self similar traffic in the context of ....

P. Jelenkovic and B. Melamed. Automated TES modeling of compressed video. In Proc. IEEE INFOCOM '95, pages 746--752, 1995.

....can generate processes with different correlation structures with uniform probability distributions. TES models use a heuristic search to find the best TES process that can capture both auto correlation structure and distribution of the empirical data. Recently, an automated solution was presented [39]. Examples of long range dependent traffic models are given in (Section 3.5) These include fractional Brownian motion, aggregate of ON OFF high variability sources, and F ARIMA. Fractional Brownian motion has only one parameter, controlling the auto correlation function. Therefore, there is no ....

B. Melamed and P. Jelenkovic, "Automated TES modeling of compressed video," in IEEE INFOCOM '95, pp. 746--752, 1995.

....the periodic statistics which have been found to be pertinent to transporting such traffic with a desired QoS[27] 28] 30] Modeling periodic MPEG video streams is even more challenging, because both the spatial (periodic) and temporal correlation need to be modeled. Several approaches[29] 16][12] have been proposed in the past. The common idea of these models was to directly model the periodic structure in the time domain. Specifically, I, P, B frames were first modeled separately, and then multiplexed in a correct order. As these methods were designed to capture spatial correlation ....

P. Jelenkovic and B. Melamed. Automated TES modeling of compressed video. In INFOCOM, pages 746--752, Toronto, Ontario, Canada, 1994.

....the model exactly captures the empirical marginal distribution (histogram) and approximately captures the empirical autocorrelation function (equivalently, spectral density) simultaneously. In addition, the fitting of parameters to a given dataset can be carried out in a systematic manner [4, 12]. While TES models have proven to be accurate in a variety of application domains [19] TES based queueing models, however, have been analytically intractable. To date, queueing models with TES based traffic have only been studied via Monte Carlo simulation. The analysis of such models is ....

....triggers a recalculation and redisplay of the corresponding statistics. TES model autocorrelations and spectral densities are calculated numerically from fast and accurate formulas developed in [8, 9] Recently, an algorithmic modeling approach has been devised and implemented for TES modeling [12], in which the TES parameters are chosen by solving a nonlinear programming problem so as to yield a weighted least squares fit between the empirical and model autocorrelation functions. 3 Discretized and Quantized TES Processes The rationale for introducing discretized and quantized TES ....

[Article contains additional citation context not shown here]

Jelenkovic, P., Melamed, B., "Automated TES Modeling of Compressed Video", Infocom '95, 746-752, 1995.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC