自從寬頻整合服務數位網路(B-ISDN)提出後，視訊傳輸已成為未來網路服
務的重點。視訊傳輸經常要求網路提供大量的頻寬，使得網路不勝負荷。
因此，視訊資料必須事先經過壓縮(譬如利用國際視訊壓縮標準MPEG或
H.261)使資料量減少，傳輸時才能降低佔用的頻寬。另外，為了避免網路
傳輸視訊時發生擁擠現象，需要一個適當的入場控制(admission
control)，來有效控制進入網路的資料量。入場控制以預測視訊傳輸的細
胞遺失機率(cell loss probability)為基準，來決定是否准許或拒絕其
建立連接(connection)。因此，首先必須建立模型來描述視訊資料量的特
性，進而分析出細胞遺失機率，最後才能完成入場控制。本論文將探討
MPEG壓縮後視訊的位元產生率行為，並且提出模型來描述其週期性現象。
藉由簡化其位元產生率圖形(bit-rate diagram)，設計出一個合成式模
型(composite model)。在模型的準確性方面，我們採用一套方法加以鑑
定，稱為注入佇列法。鑑定結果發現，對於統計值(statistic)和平均細
胞時間延遲(average cell delay)，模型皆能準確地計算出來。然而對於
細胞遺失機率，模型之準確性略差，但經過分類後，準確性獲得改進，並
且已足夠提供入場控制所需。另外，不論視訊序列中是否具有場景轉換(
scene change)或無動作狀態(motionless)，模型都能發揮作用。

Since Broadband ISDN had been proposed, video transmission has
become more significant on network service in the future. Video
transmission usually impose very large bandwidth requirements
on the network; therefore, video data are usually compressed
using international standards such as MPEG, H.261. Moreover, in
order to prevent network from congestion for video
transmission, an admission control for video traffic is
essential. Since the admission control may be based on the cell
loss probability of the channel, video models are usually used
to find out the cell loss probability in order to implement
admission control. In this paper, we aim at the behavior of bit-
rates generated by MPEG encoded packet video and intend to
propose a video model to describe the periodical phenomenon of
bit-rates. Through simplification on bit-rate diagram, a
composited model is presented. To verify the accuracy of the
proposed model, we employ a verification method called as
Feeding-Queue. The results of verification show that the
statistics and the average cell delay can be accurately
computed. As for the cell loss probability, the proposed model
is not so accurate as for the statistics and the average cell
delay. Through classification, the accuracy has been improved
and is good enough for the analysis of admission control.
Moreover, the performance of this model can be guaranteed no
matter if there are scene changes or motionless frames in the
video sequence.