Video Case

In the case of video transmission, to show the effect of changing some of the controlling parameters, we proceed as follows. Based on the experimental and simulation studies of EB-TFRC given in [41], and based on the governing equation 8.1, we suppose for simplicity that for different network states, the sending rate as suggested by EB-TFRC is as depicted by the dot-dashed line in Figure 8.4. The fluctuations in the rates are due to the effect of the different variables in that equation. This is to take into account the variations of the end-to-end delay, the packet size, the loss event, the number of competing connections, etc. As we previously mentioned, the sender should not (in some cases cannot) follow the fluctuations of TFRC, to avoid degrading the quality. In addition, for the purpose of clarifying the idea, we also suppose that the sender changes the sending rate according to the solid line depicted in the same Figure. We have selected two controlling parameters in order to meet the suggested rates, namely the quantization parameter (QP) and the frame rate (FR). By changing the QP, the encoder can compress the original signal to the level we want, and hence produce the desired bit rate (BR). (See Sections 3.4 and 6.2 for more details). By skipping some frames from the original video signal, and hence changing the FR, also the BR can be changed to meet the required sending rate. For the case of using only the QP as controlling variable, we fixed FR to 30 frames/s and RA to 0.05. In the congested periods, we suppose that there is 1% loss with CLP of 1. Similarly, for the case of FR, we fixed all the parameters as before except BR and FR. The QP in this case is the one that gives the best quality (the maximum BR). We used the trained RNN (see Chapters 6 and 7) to measure the quality for each case. We show in Figure 8.5 the quality evaluation for both cases. As we can see, the case of changing the FR gives better quality than that when allowing only the QP to change. The improvement of the quality is more important in the case when the sending rate has to be very small. The explanation of this fact has already been mentioned before. Briefly, by fixing the FR to 30 frames/s and changing only the QP to reduce the required BR, the impairments of the encoded signal increase due to the encoding artifacts. On the other hand, by fixing the QP to its best value, and reducing only the FR, the impairment level is not as annoying from the user's perception as the other case. This is because human eyes are less sensitive to the change in FR rather than luminance and chrominance distortions (see Section 7.4.2).

Figure 8.4: The supposed rates suggested by TCP-friendly and that of the sender

Figure: MOS values when changing frame rate and those when changing the quantization parameter to meet the bit rates shown in Figure 8.4