在近幾年來科技上的進步突飛猛進，尤其是在網路應用的方面更是日行千里，使得現在的網路上充滿著許多多媒體的服務，然而對這些網路狀況若是能夠事先獲得越多，那麼便能夠對系統的傳送速度作適當的調整，以利接下來的網路傳輸，
然而在收集網路狀態資訊的研究方面，大部分的判定方法皆是使用Loss Rate當作一個網路狀況的指標，然而本論文認為使用Loss Rate並不足以代表Real-Time的網路特性，因此我們使用RTP來傳輸資料封包，在RTP每收到一個封包皆可由其RTP的標頭內獲得資訊來計算封包的延遲時間，這個延遲時間我們就拿來當作我們的網路狀況的調整參考值，為了要即時的反應出網路的改變，我們使用了neural network來做即時的預測分析下一個封包可能的延遲時間，並立即針對此預測數值來調整我們的BitRate，這是我們的短程動態調整。

除短程的動態調整外，我們嘗試去將網路分成五個狀態別，用來表現出實際網路的情況，並依照所判斷出來的網路狀態來選擇傳送哪一個Layer的資料，其中的Layer可分成Enhance Layer 與Base Layer，若是落在差的網路狀況時便將Enhance拿掉不傳送，便可降低許多網路上的負載。另外我們針對影響語音品質的連續遺失封包情況以多送的方式作處理。

In recently years the development in the field of network technology grows rapidly especially for the Internet applications, thus, a variety of multimedia services currently block up the IP networks. Due to the limitation of available network bandwidth and unpredictable transmission delay, services of streaming video/audio on such unstable environment often seriously affect the quality of services. However, if we can get the network traffic information as much as possible then we will be able to adjust the delivering parameters, such as packet rate, which will benefit the succeeding service quality.
In the aspect of the network traffic information collection, loss rate is often used as the indicator in the literatures. In our experience, it is not adequate that utilizes the factor of loss rate as the representation of network traffic characteristic only. In this study, we use both the delay and jitter of packets obtaining from suitable RTCP packets. Besides, to react the change of the network condition rapidly we employ a neural network computation to predict and analyze the possible delay the ongoing packet will bring. Then, we will be able to adjust the delivering parameters accordingly.

In addition to the short-term dynamic adjustment, we represent the actual network traffic condition as five levels, whether to deliver the information or not will be decided under the concern of current level. If it is not stable, the rate of information transfer will be reduced immediately by plummeting some miner information. Finally, we adopt double-sending strategy to avoid consecutively dropping packet that will let the quality become poor.

【1】 T. Turletti，MIT and C.Huitema，Bellcore，『RTP Payload Format for H.263 Video Streams』， RFC 2032，October 1996
【2】 Audio-Video Transport Working Group，Precept Software, Inc.，R.，Xerox Palo Alto Research Center and Lawrence Berkeley National Laboratory，『RTP: A Transport Protocol for Real-Time Applications』， RFC 1889，January 1996。
【3】 Audio-Video Transport Working Group，H. Schulzrinne and GMD Fokus『RTP Profile for Audio and Video Conferences with Minimal Control 』，RFC 1890，January 1996。
【4】 Randa E1-Marakby，David Hutchison Computer Dept.，『Delivery of Real-time Continuous Media over Internet』，Computers and communication，1997，Proceeding，Second IEEE，1997。
【5】 P.C.K. Liu，W.L. Cheung and S.W.J Tam，『The Modelling of Variable Bit Rate Video Traffic』，International Symposium on Speech，Processing and Neural Networks，April 1994。
【6】 M. Reha Civanlar，AT&T Labs，Newman Springs Lab，『Protocals For Real-Time Multimedia Data Transmission Over The Internet』
【7】 Zhao DongSheng，Lu XiCheng，Zhou XingMing，『Dynamic Quality of Session Control of Real-time Video Multicast』，IEEE International Conference on Intelligent Processing Systems，October 1997
【8】 SNTP (Simple Network Time Protocol): 『RFC 1769，網路校時程式NTPClock』 http://squall.cs.ntou.edu.tw/ntpcal/download.html
【9】 Reza Rejaie，Mark Handley，Deborah Estrin ,『RAP：An End-to-End Rate-based Congestion Control Mechanism for Realtime Streams in the Internet』，INFOCOM '99。