As is mentioned in Section 1, delay jitter is a critical factor in real-time voice applications, which obstructs the proper and timely reconstruction of the stream at the receiving end. Under the stringent delay requirements, packets could get lost due to their late arrival resulting from increasing network delay. One important functionality to be implemented at the receiver is the playout scheduling of the voice packets, or in other words, setting the time when to play out the received packets. With the existence of delay jitter, the playout scheme greatly affects the tradeoff between loss and latency. In our previous work [10], an adaptive playout scheme is proposed, which dynamically adjust the playout schedule according to the varying delay jitter statistics.
With multiple streams which experience different delay, the problem of playout scheduling becomes more complicated, with one more concern on the quality of the reconstructed speech. In the next subsection, we will discuss the encoding of the source signal and how it affects the quality of received speech. In Subsection 2.3, we will present our playout scheduling algorithm for multiple streams.