110 likes | 254 Vues
Dr. Abhay Karandikar from IIT Bombay discusses the characteristics and challenges of Voice Over Internet Protocol (VoIP). He explains key issues including audio generation, talk spurts, packet loss tolerance, and delay management. The impact of packet loss on real-time services is evaluated, along with techniques such as Forward Error Correction (FEC). Additionally, he covers adaptive playback algorithms for handling jitter and estimating network delays to improve voice quality. This overview highlights critical considerations for enhancing VoIP performance.
E N D
Voice Over Internet- Issues Dr. Abhay Karandikar Department of Electrical Engineering Indian Institute of Technology, Bombay. Dr. Abhay Karandikar Department of Electrical Engineering Indian Institute of Technology, Bombay.
Characteristics of Audio / Voice over Internet • Internet phone generates audio consisting of talk spurts and silent periods • Talk Spurt - 8 Kbytes/sec • Bytes into chunks every 20 msec, header is appended • UDP segment is sent
Problem for Real Time Services • Packet Loss • Retransmission mechanism is not suitable • 1% to 20% packet loss can be tolerated depending upon application • FEC can help conceal Packet Loss • End To End Delay • Delay smaller than 150ms are not perceived by human • 150 - 400 ms delay acceptable • Delay jitter • Due to random queuing delays
Handling Jitter at the Receiver • Append ‘Sequence Number’ • Append ‘Time Stamp’ • Use ‘Playback Buffer’
Fixed Playout Delay Graph Packet Generation N/W delay Buffer Sequence Number Playback Time
Adaptive Playback Buffer • Estimate the network delay and the variance of the delay • Adjust the playback delay at the beginning of each talk spurt • May cause sender’s silence period to be compressed and elongated
Adaptive Playback Algorithm for Internet Phone Let ti = timestamp of the i packet ri = time of arrival of ith packet at the receiver pi = Playback time of the ith packet di = Estimated Average network delay upon reception of ith packet vi = Average deviation of the delay from the estimated average delay
continued…. di and vi are updated as follows. di = ( 1-a )di-1 + a( ri - ti ) vi = ( 1-a )vi-1 + a | ri - ti - di | Playback time is calculated as pi = ti + di + bvi
Playout Time of the next Packet in the Spurt If ith is the first packet of the talk spurt Let qi = ti - pi then pj = tj + qi
Problem • How to determine first packet of a talk spurt? If ti - ti-1 > 20 ms then ith packet is first packet of the talk spurt • What if there is a packet loss? Take the help of Sequence Number
Issues • Question • Who does • Appending Sequence Number? • Timestamping to audio/video chunks? • Answer • Real-time Transport Protocol