Media Compression - Scalable Continuous Media Streaming Systems - page 29

audio bit-stream (audio packets)

Audio
Input

Audio
Encoder

Multiplexer
(System Encoder)

System bit-stream
(system packs)

Video
Encoder

Video
Input

video bit-stream (video packets)

Presentation Time Stamp
(PTS) Added

Decoding Time Stamp
(DTS) Added

Figure 2.3 The encoding process in a typical MPEG media encoder

system stream and multiplexes the audio and video data streams according to their timing cor-
relations. This system stream can then be delivered over the network to the clients for playback,
where the reverse decoding process will occur, i.e., it first demultiplexes the system stream into
separate audio and video streams, and then decompresses them for synchronized playback.
In media streaming, the media multiplexer can reduce the complexity of the media server.
In fact, the media server can simply treat the multiplexed system stream as a binary bit stream
encoded at a combined system stream bit-rate, irrespective of the detail compression algorithms
and data format employed. This is a significant advantage as it decouples the media server
implementation from the media compression standard employed. In other words, we can reuse
without modification the same media server to stream media data compressed using new
compression algorithms when they become available. By the same token, we can also stream
multiple types of media streams encoded with different compression standards using the same
media server, thereby reducing cost and operational complexity.
Alternatively, a media server can also send the compressed audio and compressed video
data streams separately over the network (cf. Chapter 6), bypassing the MPEG multiplexer
altogether. In this case, the media server will then need to send the data streams in such a way
that audio and video data will arrive in time for synchronized playback. This usually requires
the media server not only to perform I/O, but also to inspect and interpret the contents of the
media streams to extract timing information to schedule data transmissions, and to construct
packet headers with presentation timing information.
In this model the media server implementation will be coupled to the media compression
standards employed, and it will also consume more processor cycles in processing the media
data. Nevertheless, this approach does give the media server more control over the data delivery
process. For example, the service provider could produce multiple versions of the media streams
at different bit-rates to cater for varying network bandwidth availability. In this case the media
server can begin streaming the highest quality media streams to a client, and then dynamically
switch to a lower quality (and thus lower bit-rate) media stream if the available network
bandwidth drops below a threshold to ensure continuous media playback. Network bandwidth
variations are common in the current Internet and thus the capability to dynamically adapt the
video content to avoid playback interruptions is an important and useful feature to the end
users (cf. Section 2.5).