next up previous contents index
Next: Simulation Results Up: Our Proposed Rate Control Previous: Our Proposed Rate Control   Contents   Index


The Possible Controlling Parameters

The following parameters can participate in the rate control protocol (refer to Chapter 7 for more details):
The bit rate (BR).
The sender can change dynamically the BR, when a variable BR (VBR) encoder is used, using multilayer encoding, and/or using another encoding algorithm. Multilayer encoding as a rate control mechanism is one of the most successful mechanisms in the literature [106,107,109,110].
Changing the codec.
Eventually, the codec type or some of the encoder's specific parameters (sampling rate, sample size, quantization parameters, etc.) can be changed to fit the needed bit rate. As described in Sections 3.4 and 3.5, there are lossless, lossy and hybrid encoding techniques. For the same bit rate and the network state, you can select the codec (or change the parameters of the current one) to get the best possible perceived quality.
The frame rate (FR).
In the case of video as experiments and study on HVS showed, the human eye is not very sensitive to variation of FR higher that 16 frames/s. Thus, by skipping some frames, the output BR can fit the value suggested by TFRC and not degrade the quality too much. This case is extensively studied in the literature as it is a very effective way of the rate control [42,127,78,76].
Forward Error Correction (FEC).
FEC can be used to protect against loss. Thus, by choosing the amount and type of redundant data [20,18,19,21,22,116], the effect of loss can be reduced and hence the quality can be improved. But too much FEC increases the overall BR and delay. Thus, a compromise should be made.
Buffer size.
Changing the buffer size in the case of real-time streaming multimedia flows, the client fetches some data from the source before starting playing the media [116]. After that, the client fetches the data whenever the size of the data in the buffer reaches some predefined value or when there is loss. By changing the size of this threshold dynamically as a function of the network state and the perceived quality, it is possible to keep some data always available to the decoder. By using our control mechanism, we believe that the quality can be enhanced working on this parameter.
Packetization interval.
In the case of speech, changing the packetization interval (and hence the packet size), the possible sending rate as suggested by the EB-TFRC will change (as it is directly proportional to the packet size from Eqn. 8.1). But, on the other hand, a too large packet size will increase the overall delay, which is not tolerable in the majority of real-time multimedia applications. In addition, it will cause the problem of fragmentation (packets greater than predefined length are divided into several packets to be fit in this length) between routers. Fragmentation increases the probability of loss because if one fragment of the packet is lost, all the fragments of that packet become useless and are hence deliberately dropped. On the other hand, larger packet size may improve the network utilization (see Section 5.2.1).

next up previous contents index
Next: Simulation Results Up: Our Proposed Rate Control Previous: Our Proposed Rate Control   Contents   Index
Samir Mohamed 2003-01-08