TITRE Adaptive Packet Video Streaming Over IP Networks - LaBRI
TITRE Adaptive Packet Video Streaming Over IP Networks - LaBRI
TITRE Adaptive Packet Video Streaming Over IP Networks - LaBRI
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
the same value. This timing information is obtained only when AUs are encapsulated in RTP<br />
packet. AUs are then transmitted to the RTP layer by the compression layer, with indication of the<br />
boundary, the payload type and other control information.<br />
Among the open issues related to MPEG-4 encapsulation and fragmentation, we have<br />
identified in the previous section, the grouping and the multiplexing issues. Hence, an appropriate<br />
ES packetization and grouping (aggregation) payload is essential for an optimal transport of<br />
MPEG-4 streams over the <strong>IP</strong> networks. The packetization process must address the overhead<br />
reduction and provides a generic RTP payload format, which offers some of the Sync Layer<br />
features. Furthermore, an efficient ESs multiplexing mechanism is much suitable for transporting<br />
multiples MPEG-4 ESs in a single RTP session such as low bit rate audio. This will facilitate the<br />
ESs management, optimize the RTP payload use, and favor the MPEG-4 terminal scalability.<br />
In this subsection, we present a new RTP payload format for MPEG-4 audio-visual object,<br />
namely, RTP4mux protocol. RTP4mux is based on ESs aggregation and offers both efficiency and<br />
robustness for MPEG-4 transport over the low bit rate <strong>IP</strong> networks such as WLAN (Wireless<br />
Local Area Network).<br />
The main problem that an encapsulation scheme must address is the overhead when<br />
transporting small AU size. Table 4-4 illustrates the low bit rate MPEG-4 streams transport<br />
overhead over RTP. It is clear that the AUs concatenation and grouping reduces the total<br />
packetization overhead.<br />
Audio Object Audio Payload Length <strong>Over</strong>head (RTP/UDP/<strong>IP</strong>) <strong>Over</strong>head Rate<br />
AAC (64 kbit/s, 24 kHz) 342 bytes (average) 12+8+20 bytes 10,9%<br />
CELP (6 kbit/s, 8 kHz) 15 bytes (fixed) 12+8+20 bytes 72,7%<br />
Table 4-4: <strong>Over</strong>head of AAC and CELP audio streams<br />
When AAC is used for encoding of a stereo audio signal at 64 kbit/s, AAC frames contain an<br />
average of approximately 342 bytes (with 1024 sample/frame). On a network with 1500 bytes MTU<br />
(Maximum Transfer Unit), a four AAC frames can be carried in one <strong>IP</strong> packet. This allows<br />
optimizing network resources and bandwidth usage. However, with the low bit rate MPEG-4<br />
streams (e.g. AAC, CELP, Facial Animation, etc.), the interval between a successive AUs generation<br />
may be important. Otherwise, the concatenation of multiple AUs from a single elementary stream<br />
will introduce a supplementary delay, which is intolerable in a real-time and interactive multimedia<br />
communications. For example, the concatenation of 30 CELP AUs will induce 600 ms<br />
packetization delay. In addition, since the low bit rate MPEG-4 streams is expected to be largely<br />
deployed on the <strong>IP</strong> networks, optimizing the bandwidth usage trough a good RTP payload<br />
exploitation is indispensable.<br />
Within RTP4mux, we propose to aggregate several MPEG-4 elementary streams on a single<br />
RTP session. We achieve a concatenation of several AUs, from several ESs, in the same RTP<br />
payload (i.e. It is possible to convey several ESs in the same RTP packet). The motivation for<br />
packetizing multiple AUs in the same RTP packet is the overhead reduction. While the motivation<br />
for concatenating AUs from different ESs is minimizing the end-to-end ESs transmission delays.<br />
Furthermore, this aggregation scheme minimizes the dependency between adjacent RTP packets,<br />
79
Change language