Challenges of Multimedia Transmission

Some of the challenges of carrying multimedia traffic over the Internet are: the lack of guarantee in terms of the bandwidth, packet loss, delay, and jitter, which affect the quality [60]. In this Section, we discuss the major problems caused by the network. $\bullet$

• Packet Loss - Packets are sent from the source to the destination across several routers. Each router may get packet streams from many sources at the same time. When the packet arrival process fills the buffer of any of the routers, some packets are dropped. The network in this case is said to be congested. Packet loss may produce great damage to the received multimedia signal. For example, for the case of speech, each packet contains 40-80 ms of speech information to match the duration of critical units of speech called phonemes. When a packet is lost, a phonemes is lost in continuous speech. While the human brain is capable of reconstructing a few lost phonemes in speech, too many packet losses make a voice unintelligible. The same problem exists for video, some portions (blocks) of the image can not be decoded and displayed. The result is severe degradation of the reconstructed signal quality. The problem becomes more severe when the signal is compressed too much. Natural signal contains in general too much redundant information that can be interpolated by human brain when there is loss. However, to reduce the bitrate, the signal is compressed to remove the redundancy, and thus, packet loss becomes more annoying. There are two types of techniques to mask the effects of packet loss: some focus on reducing packet loss, others concentrate on repairing the damage caused by loss. To reduce packet loss; Network upgrade (increasing the bandwidth, using ATM, SONET [Synchronous Optical NETwork] for gigabits/s, Wavelength Division Multiplexing [WDM] for terabit/s) is needed. This solution is very expensive and does always available. We overview the other methods dealing with packet loss in the next Subsection.
• Packet Delay - To understand the effect of packet delay on multimedia quality, we take the case of voice transmission. Timing is an important characteristics of voice. Two syllables of a word are uttered with an interval. This interval is as much a part of the voice as the uttered syllable. If additional delay is inserted between syllables, the rhythm of voice is lost. Long delay may force conversation to be half-duplex or introduce echo. Delay below 150 ms is acceptable. For long distance communication delay between 150 and 400 ms is also acceptable. The end-to-end delay typically consists of: codec delay, serialization delay (the time it takes to place a packet in the transmission line), queuing delay (in routers, etc.), propagation delay (time to travel from point to another). From that it is clear that to reduce the delay effect we have several possibilities: (i) using codecs that can run in real time without too much delay, (ii) reduce the serialization delay, (iii) increase the routers speed to decrease the queuing delay or using some kind of Differentiated Services, (iv) reduce the length of the phisical material to reduce the propagation delay (this can be done by selecting the shortest pass from the source to the destination in the network layer). It should be mentioned that delay effect is not relevant in the case of one-way sessions (video streaming, etc.). However, the effect of delay play an important role in the case of interactive two-way sessions [60].
• Network Jitter - It is the variance of the inter-packet arrival time at the receiver. Network jitter occurs due to the variability in queuing and propagation delays. To alleviate from jitter, a jitter buffer is used. The receiver holds the first packet in the buffer for a while before playing it out. The amount of the hold time is the measure of the size of the jitter buffer. For example, a 50 ms hold time means 50 ms jitter buffer. Jitter, on the other hand, plays an important role in both types of applications (interactive or one-way), and its effects are similar to those of packet loss. If a dejittering buffer [37] is implemented, the effect of jitter is reduced, and from the point of view of the application, the effect of jitter can be translated as extra network losses. As packets that arrive after some expiration time are considered lost. Finally, for interactive applications, the echo effect should also be considered, and an echo suppression or cancellation mechanism should be implemented [40].