Logo image
串流語音之延遲變異減緩與封包遺失減少演算法
Thesis

串流語音之延遲變異減緩與封包遺失減少演算法

吳晉愷
Masters, 國立清華大學, 資訊工程學系
2001

Abstract

延遲變異 封包遺失 減少 串流 語音 Jitter Packet Loss Reduction Streaming Voice
The vast progresses in Internet and multimedia related technologies have brought dramatic and fundamental changes to the world that we live. Recently, the emerging high quality streaming audio, voice, and video for multimedia services on the Internet is progressively popular; voice over IP (VoIP) and video on demand (VOD) services are two typical examples. Thus, owing to the rapidly growing of bandwidth on IP networks, Internet-telephony that delivers a large amount of compressed voice IP packets to optimize the bandwidth utilization becomes one of the most favorite services. Internet-telephony is admired not only its cheaper to build the infrastructure, moreover, it can provide various value-added services in the future. On certain streaming services, it is also significant to provide guaranteed QoS (quality of services) to satisfy the real-time and stringent delay requirement because the real networks are heterogeneous in bandwidth and reliability. Guaranteed QoS includes three basic parameters: packet transfer delay, packet delay variation and packet loss rate. Once the voice packets sent from the source end, the variation of networks makes the condition of voice packets unpredictable. As a result, there are many researches concerning providing guaranteed QoS schemes. Early related works tend to provide QoS at the source end or the destination end with interactive feedback mechanism. Recent works, such as jitter earliest due date queueing (JEDD) and multilayer-gated frame queueing (MGFQ) tries to achieve the goal in the intermediate nodes. In this thesis, two on-line algorithms for providing QoS tuning at the intermediate nodes are proposed. First, the jitter control algorithm aims at smoothing the jitter variance with low queueing delay latency. So that the buffer allocated at destination becomes smaller. Second, the burst packet loss reduction algorithm is designated to decrease the frequency of continuous packet loss. This makes the voice stream more robust to go well with burst or congestion conditions, thus, can be performed with higher quality at the client site.

Metrics

1 Record Views

Details

Logo image