An MPEG-2 video traffic prediction based on phase space analysis and its application to on-line dynamic bandwidth allocation

Real-Time VBR Video Traffic Prediction for Dynamic Bandwidth Allocation

“…Due to the rapid development of a new paradigm of intelligent systems and methodology called computational intelligence (including fuzzy logic, neural networks and evolutionary computing), many novel algorithms and applications of computational intelligence in telecommunications systems including various aspects of ATM networks (such as connection admission control, policing, and routing) have been proposed and developed [e.g., [11]- [15]]. However, in the recently published works on direct traffic prediction using various techniques including fuzzy logic (e.g., [16]), neural networks (e.g., [5], [17]- [19]), and adaptive linear models (e.g., [20]- [22]), studies were only conducted on one-frame-ahead traffic prediction. One-frame-ahead traffic prediction, i.e., predicting only next one video transmission frame at a time, is a very short-term traffic prediction.…”

Sparse Basis Selection: New Results and Application to Adaptive Prediction of Video Source Traffic

“…Also, assume that new measurements are now available at time , and that it is desirable to update the set and its corresponding weight vector . To simplify the notation, let (8) (9) To make the method adaptive and have the recent data have a larger influence on the selected vectors, a discounted error function is used (10) where is the th row of , and is the discount weight, in the range from 0 to 1 (usually closer to 1).…”

“…The inputs for the P-frame prediction model are as follows: [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16] the same as for the previous I-frames but applied on P-frames 17) previous P-frame minus mean of P-frames 18) difference between previous 2 P-frames 19) mean of previous 2 P-frames 20)…”

“…The inputs for the B-frame model were as follows: [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16] the same as for the previous I-frames but applied on B-frames 17) previous P-frame minus mean of P-frames 18) mean of previous 2 P-frames 19) mean of previous 3 P-frames 20) mean of previous (B1, B2) for B1 and B2 frame types, similar input for other B-frame types 21) mean of (B5, B6) for B1 and B2 frames and vice versa, mean of (B7, B8) for B3 and B4 frames and vice versa 22) previous I-frame minus mean of I-frames 23) mean of previous 2 B-frames 24) mean of previous 2 I-frames logarithm of inputs (1)- (24) appropriately scaled 49-72) exponential of inputs (1)- (24) appropriately scaled 73-96) square root of inputs (1)- (24) appropriately scaled 97-120) sigmoid of inputs (1)- (24) appropriately scaled.…”

“…Both Adas's and Yoo's approaches are adaptive, in the sense that they continually update the predictor weight coefficients with time. Chodorek and Chodorek [10] develop a linear predictor for MPEG-coded video traffic based on partitioning of the phase-space into subregions.…”