Jean-Chrysostome Bolot and Andrs Vega Garca, "Control mechanisms for packet audio in the Internet," Proc. IEEE INFOCOM, 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....K=20 K=50 K=1000 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Amount of FEC a r = 0.9 K=5 K=10 K=20 K=50 K=1000 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Amount of FEC a r = 1.5 K=5 K=10 K=20 K=50 K=1000 Fig. 3. OE = 1 and the queue capacity is not changed. related [16], 17] 18] Here, we are interested in finding the probability that packet n OE is lost given that packet n is also lost. This will give us P (Y n OE = 1jY n = 0) which in turn gives us the expression for the audio quality (as expressed in (3) Since we assume that the system is in its ....

....that audio quality always deteriorates when applying this kind of FEC mechanism. It is therefore desirable to study other FEC methods that can provide a better quality. Recently, Ratton [19] found that media independent FEC techniques using parity bits ( 5] perform better than media specific FEC [16], 17] 5] We can provide an intuitive explanation to the reason that the simplistic FEC studied here does not perform well. In this scheme, each added unit of redundancy protects only one unit of information that can be retrieved. There is only one possibility to retrieve a lost packet. We can ....

Jean-Chrysostome Bolot and Andrs Vega Garca, "Control mechanisms for packet audio in the Internet," Proc. IEEE INFOCOM, 1996.

....media content and transmit it via the network as packets at regular intervals. The receiver gets the media packets and schedules an appropriate playout time in order to produce a smooth output media stream. It compensates for the delay variation (jitter) using a playout delay adjustment algorithm [7, 16, 14, 3, 2]. Simple algorithms use a xed playout delay, either static or determined at the start of a session. More advanced VoIP applications compute a di erent playout delay for each talk spurt [4] adaptively according to the current network condition. The quality of multimedia applications is primarily ....

Jean-Chrysostome Bolot and Andres Vega Garcia. Control mechanisms for packet audio in the internet. In Proceedings of the Conference on Computer Communications (IEEE Infocom), San Fransisco, California, March 1996.

....K=50 K=1000 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Audio quality Q(a) Amount of FEC a r = 0.9 K=5 K=10 K=20 K=50 K=1000 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Audio quality Q(a) Amount of FEC a r = 1.5 K=5 K=10 K=20 K=50 K=1000 Fig. 3. and the queue capacity is not changed. related [16], 17] 18] Here, we are interested in finding the probability that packet is lost given that packet is also lost. This will give us which in turn gives us the expression for the audio quality (as expressed in (3) Since we assume that the system is in its ....

....that audio quality always deteriorates when applying this kind of FEC mechanism. It is therefore desirable to study other FEC methods that can provide a better quality. Recently, Ratton [19] found that media independent FEC techniques using parity bits ( 5] perform better than media specific FEC [16], 17] 5] We can provide an intuitive explanation to the reason that the simplistic FEC studied here does not perform well. In this scheme, each added unit of redundancy protects only one unit of information that can be retrieved. There is only one possibility to retrieve a lost packet. We can ....

Jean-Chrysostome Bolot and Andrs Vega Garca, "Control mechanisms for packet audio in the Internet," Proc. IEEE INFOCOM, 1996.

....configuration, i.e. they should be able to determine the optimal way of distributing themselves and the distribution may have to be changed during execution. The problem of adaptation has received a fair bit of attention, especially in the area of multimedia, e.g. network aware streaming of audio [6, 2] and video [5, 14, 7, 9] data. The performance of distributed computing applications tends to be much harder to characterize, making it more di#cult to make them network aware. In this paper we introduce two di#erent mechanisms for making an application network aware, one based on a performance ....

....that do congestion avoidance (e.g. TCP IP [8] are an important example of adaptation to network conditions. In terms of applicationspecific adaptation to network conditions by network aware applications, several groups have focused on continuous media tra#c streams, e.g. network aware voice [6, 2] and video [5, 14, 7, 9, 10] streaming applications. There has been some work on non multimedia network aware applications, although the network aware decision making is typically at a coarse level. One example is having storage intensive applications (e.g. Web browsers) select the most ....

Jean-Chrysostome Bolot and Andres Vega-Garcia. Control mechanisms for packet audio in the internet. In IEEE INFOCOM'96, volume 1, pages 232--239, San Francisco, CA, March 1996. IEEE.

....to adapt based on implicit feedback, e.g. TCP interprets dropped packets as a sign of congestion [19] Recently, several applications have been developed 15 that adapt to the bandwidth and latency variations of the underlying network. One class of examples are Internet based video and audio tools [17, 18, 15, 6, 31]. Another class of examples consists of distributed computations modifying the granularity of the computation in response to network status [27, 29] The adaptation policies used in these applications are ad hoc and application specific. Interpreting implicit feedback from the network is ....

Jean-Chrysostome Bolot and Andres Vega-Garcia. Control mechanisms for packet audio in the internet. In IEEE INFOCOM'96, volume 1, pages 232--239, San Francisco, CA, March 1996. IEEE.

....of all packets form a line y = x 13. An alternative de nition of playout delay is the delay between sending time and playout time. The reader can choose either of the two de nitions as long as its meaning is clear. Many techniques have been developed for controlling playout delay [13] 36] 28] [7] [6] etc. Simple ones use a xed playout delay, either static or determined at the start of a session. More advanced techniques exploit the existence of talk spurts (speech) and pauses (silence) 8] in human speech. The length distribution of talk spurts and pauses depends on silence detector ....

Jean-Chrysostome Bolot and Andres Vega Garcia. Control mechanisms for packet audio in the internet. In Proceedings of the Conference on Computer Communications (IEEE Infocom), San Fransisco, California, March 1996.

....of all packets form a line # = # 13. An alternative de nition of playout delay is the delaybetween sending time and playout time. The reader can choose either of the two de nitions as long as its meaning is clear. Many techniques have been developed for controlling playout delay [13] 36] 28][7][6] etc. Simple ones use a xed playout delay, either static or determined at the start of a session. More advanced techniques exploit the existence of talk spurts (speech) and pauses (silence) 8]inhuman speech. The length distribution of talk spurts and pauses depends on silence detector ....

Jean-Chrysostome Bolot and Andres Vega Garcia. Control mechanisms for packet audio in the internet. In ########### ## ### ########## ## ######## ############## ##### ########, San Fransisco, California, March 1996.

....of missing data using the surrounding packets) and interleaving. There has been much interest in the use of packet level FEC for sending redundant information ahead of time to compensate for loss, based on parity codes, 9] 10] Reed Solomon codes [11] and redundant speech codecs [12] 13] [14] [15] Oftentimes, the term FEC is only applied to the traditional channel coding approaches, such as parity and Reed Solomon codes. For purposes of this paper, we define FEC as any mechanism which sends additional information along with the media stream, the purpose of which is to aid in packet ....

....described in [18] 19] fall into this category. The dependency of pR on the delay distribution and loss probabilities becomes more complex when FEC is introduced. The result is that the choice of D is more strongly influenced by p. Consider the FEC mechanisms described by Bolot and Garcia [14]. These algorithms use multiple low bitrate versions of a packet, each version being piggybacked on a subsequent packet. A packet is played out so long as a receiver gets the original or any one of the K 1 redundant versions of the packet on time. If packets are lost independently with ....

Jean-Chrysostome Bolot and Andres Vega Garcia, "Control mechanisms for packet audio in the internet," in Proceedings of the Conference on Computer Communications (IEEE Infocom), San Fransisco, California, Mar. 1996.

....which facilitates the creation of efficient schedules for dynamic striping. 4.2.3 Forward Error Correction Recent studies of the MBone[33, 34, 63] reveal that reception of a packet at all receivers is highly unlikely. We must make our protocol work for networks with these characteristics[29]. The experiments conducted by Yajnik[166] found the probability of a packet reaching all destinations to be 45.1 on average for MBone applications such as Radio Free VAT with a small (8 12) number of receivers. In studying applications such as the NASA Shuttle Video session with much larger ....

....of loss virtually guarantees retransmission of every packet for large groups that require reliable data delivery. The extremely high probability of loss in multicast transmission has led many researchers to propose the use of Forward Error Correction (FEC) mechanisms to assist in reliable delivery[29, 76, 122, 150]. As multicast becomes ubiquitous in the Internet, the high error rates are expected to abate somewhat; however, even lower loss rates will most likely benefit from some form of FEC. 4.2.3.1 Basic Operation of FEC In Forward Error Correction, the sender transmits redundant information so that ....

[Article contains additional citation context not shown here]

Jean-Chrysostome Bolot and Andres Vega-Garcia. Control mechanisms for packet audio in the internet. In Proceedings of INFOCOM'96, pages 232--239. IEEE, April 1996.

....Fall, 1998] The aim of an application transmitting a real time stream should be to detect congestion, through feedback information from receivers, and react by reducing the transmission rate. There have recently been some attempts at providing network friendly real time audio applications. In [Bolot and Garcia, 1996] , an adaptation scheme for multicast conferencing is described, where the transmission parameters are modified, based on averaged group statistics, collected through RTCP receiver reports. This is achievable 3.2. Robust Audio Tool Solutions 48 with unicast communication (two way conferencing) ....

Jean-Chrysostome Bolot and Andres Vega Garcia. Control mechanisms for packet audio in the internet. In Conference on Computer Communications (IEEE Infocom), San Fransisco, California, March 1996.

....The same sort of receiver autonomy is needed to ensure reliable delivery, otherwise the sender will be overwhelmed with error control information. Researchers have proposed the use of Forward Error Correction (FEC) to recover from packet loss at the receiver in reliable multicast protocols[5], 13] 14] 17] 16] A (k; n) encoder takes k packets of source data and produces n Gamma k redundancy packets such that any k of the n (k original n Gamma k redundancy) packets can be used to reconstruct the original k packets[10] 15] The drawback to FEC is the complexity of encoding ....

Jean-Chrysostome Bolot and Andres Vega-Garcia. Control mechanisms for packet audio in the internet. In Proceedings of INFOCOM'96, pages 232--239. IEEE, April 1996.

....advantages, there are a number of barriers to more widespread use of Internet telephony. Most prominent among them is the poor quality of voice connections on the Internet [5] 6] 7] This problem is being attacked on a number of fronts, including forward error correction to recover from loss [8] [9] adaptive playout buffers [10] 11] for jitter absorption, and resource reservation for improved network QoS [12] 13] Put togetther, all of these should improve voice quality on the Internet. This is only half the picture, however. Internet telephony is only as useful as the set of people ....

Jean-Chrysostome Bolot and Andres Vega Garcia, "Control mechanisms for packet audio in the internet," in Proceedings of the Conference on Computer Communications (IEEE Infocom), San Fransisco, California, Mar. 1996.

....packet B must listen again to packet A before receiving packet B. The time to receive the extraneous packet A results in inefficient use of the channel. Researchers have proposed the use of Forward Error Correction (FEC) to recover from packet loss at the receiver in reliable multicast protocols[5, 9, 15, 16, 20, 19]. A (k; n) encoder takes k packets of source data and produces n Gamma k redundancy packets such that any k of the n (k original n Gamma k redundancy) packets can be used to reconstruct the original k packets[11, 17, 18] Consider the example in Figure 2 where the sender creates two redundancy ....

Jean-Chrysostome Bolot and Andres Vega-Garcia. Control mechanisms for packet audio in the internet. In Proceedings of INFOCOM'96, pages 232--239. IEEE, April 1996.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC