Paper Review: Adaptive Playout Mechanisms for Packetized Audio Applications in Wide-Area Networks
Reviewer: Kenneth Chin
The main idea in the paper is measuring and analysing 4 adaptive playout mechanisms as to how each of them is benefitial to packetized audio applications in WAN. Basically, the adaptive playout mentioned is delayed playout. However, it is good for us to know how much delay is enough before the buffered packets are played. Obviously, it makes no sense to buffer the whole audio before it is played out, especially when the audio is long. One idea is that if we can buffer some packets ahead the playout, we can hide users from the late packets arrival due to network delays. The 4 mechanisms mentioned in the paper essentially exploit the same idea; for every packet, calculate the time from a packet is sent until it is played out, then take the mean of the previous calculations. It is found that the 4th mechanism is the best adaptive playout mechanism because it takes care of the spike anormally occured in the Internet.
All the adaptive playout mechanisms dicussed in the paper bear the same potential problem which is that they assume the clock in both sender and reciever are somehow synchronized. It is a very vulnerable assumption!
One thing I learn from this paper is that people are taking advantage of the silent period between successive talkspurts. It is a very good approach for speeches, but it might not be good for continuous audio such as music.
This paper does give a flavor of how adaptive playout mechanisms differ from each other and from the results, it is clear that a lot of tuning can be done in the future. Therefore, it should be a 1st grade paper, meaning that it gives incremental contribution.