Adaptive Delay aware error control for Internet Telephony
Real-time audio over Best-Effort networks often suffers from varying packet loss rates, delays and available bandwidth. Forward Error Correction (FEC) is an efficient way to cope with packet losses. But the use of FEC has one main drawback: it increases the end-to-end delay (the more redundant information is sent, the longer the destination has to wait to receive it). Yet, it is recognized that above a certain threshold (around 150ms), the end-to-end delay becomes noticeable and may have a non negligible impact on the perceived audio quality. Existing error control schemes for audio aim at maximizing the audio quality at the destination without taking the end-to-end delay into account. This often leads to delays unnecessarily passing this critical threshold. In this paper, we develop an adaptive error control scheme for audio which is delay aware, namely, which chooses the FEC according to its impact on the end-to-end delay. To this end, we model the perceived audio quality as a function of the encoding rate received at the destination and the end-to-end delay. Then, we develop joint rate/error/delay control algorithms which: (1) optimize this measure of audio quality (also called utility) and (2) are TCP-Friendly. The validation of a particular model of perceived quality is out the scope of this paper. In a single class best effort network (also refered as 'Flat' network), we show that our scheme increases the utility by avoiding that a source waste delay on FEC when it is not necessary. Then, we also consider Alternative Best Effort (ABE) networks. ABE is a differentiated service that offers Applications the trade-off between receiving a lower end-to-end delay or more overall throughput. We investigate whether there is a real benefit, for the audio sources, to trade delay for throughput (and packet losses).
Record created on 2005-07-13, modified on 2016-08-08