REDUP: a packet loss recovery scheme for real-time audio streaming over wireless IP networks

Due to the characteristics of (1) smaller bandwidth and (2) unreliable transmission media, real-time audio streaming over wireless networks is not trivial. To have smooth audio streaming over wireless networks, we propose a scheme called REDUP in this paper. Two sending modes, in which redundant information is embedded in each packet, that the REDUP scheme contains are (1) the ‘‘redundant’’ mode and (2) the ‘‘duplicated’’ mode. Let a packet i can contain three audio frames i, i 1, and i 2. In the redundant mode, frame i uses a codec of better quality than that for frames i 1 and i 2. In the duplicated mode, frames i, i 1, and i 2 use the same codec, which has lower quality than that for frame i used in the redundant mode. The ‘‘redundant’’ mode may give better quality of sound but consumes more bandwidth, while the ‘‘duplicated’’ mode gives lower quality of sound but consumes less bandwidth. The two modes are selected depending on the networking situation. Round trip time (RTT) between the mobile host and the mobile gateway is used to determine the networking situation. Two thresholds named ‘‘upper-ratio’’ and ‘‘lower-ratio’’ are set. When the average RTT exceeds the upper-ratio multiplies the maximum RTT, the networking situation is set to congested; when the average RTT is under the lower-ratio multiplies the maximum RTT, the networking situation is set to unloaded. The transmission mode is switched from ‘‘redundant’’ to ‘‘duplicated’’ when the networking situation is determined to be congested. The transmission mode is switched from ‘‘duplicated’’ to ‘‘redundant’’ when the networking situation is determined to be unloaded. In this way, the redundant information of packets is adapted to the networking situation and the bandwidth can be utilized more effectively. 2004 Elsevier Inc. All rights reserved.