Please use this identifier to cite or link to this item:
|Title: ||802.16e 網路之語音串流性能分析|
Speech activity detection (SAD)
Packet dropping probability.
|Issue Date: ||2010-05-13 11:56:09 (UTC+8)|
|Abstract: ||本計畫探討語音串流於802.16e 網路的性能，當語音源處於靜音(Silence)的狀態下無封包產生，而當處於講話(Talkspurt)狀態時週期地產生封包，此一語音活動性偵測(Speech activity detection,SAD)機制是由一個二狀態的馬可夫鏈描述。在不違反現行已定的標準下，我們建議使用上行鏈路中generic MAC header 中的兩個保留位元來給Subscriber Station (SS)傳達所服務的語音連接中狀態的訊息至Base Station(BS)，BS 據此增加或減少分給SS 的上行頻寬。在此操作下語音串流就如同Extended real-time Polling Service (ertPS)，當SS 所有的語音源都處於靜音時，因無封包傳送於上行頻寬，我們建議改採rtPS 服務，亦即BS 週期的詢問SS 直至至月一語音源轉換到講話狀態，因而我們的提議就如同結合ertPS 和rtPS 服務。為了比較我們也會探討當SS 由無講話音源改變至至少有一講話時改採競爭模式，此模式可能造成競爭碰撞而產生延遲，但可避免輪詢造成的頻寬消耗，所以我們將比較rtPS 模式和競爭模式的性能差異以求取系統的最佳性能。在另外一方面，在TDD 結構下的訊框架中下鏈及上鏈次訊框邊界是可動態調整，這特性極為適合具SAD 的語音串流，因而我們也將探討機動調整次訊框邊界所能帶來的性能優點。系統性能的指標為封包遺漏機率(Packet dropping probability)、接取延遲(Access delay)及通道流通量(Channel throughput)。此外VoIP 支援不同編碼率的語音源，我們也將探討具有不同編碼率語音源的系統性能。
This project studies the performance of voice streams over 802.16e networks. We assume the voice source employs a speech activity detection (SAD) mechanism so no packets are generated when the source stays in silence state and packets are periodically generated when it stays in talkspurt state. The SAD mechanism can be described by a two-dimensional Markov chain. Without violating what have been standardized, we suggest to use the two reserved bits in uplink generic MAC header of each MAC protocol data unit (MPDU) for the subscriber station (SS) to inform the base station (BS) about the state changes in its voice connections. The BS then adjusts the uplink bandwidths accordingly. Operating in this way, the voice streams may be viewed as the extended real-time polling service (ertPS) class. When all the connections of an SS are in silence state, as no uplink bandwidth available, the BS periodically polls the SS until a bandwidth request is received. Under this situation, the voice streams look like a rtPS class. Thus, our proposal is a combination of ertPS and rtPS mode. For comparison purposes, we also examine the performance of using contention opportunities when the SS’s state changes from no talkspur to at least one talkspurt. In this way, although the usage of bandwidths for polling is avoided, contentions may result in collisions, and thus introduce additional access delay. Thus, we are interested in the performance difference between rtPS and contention modes. On the other hand, as the TDD frame structure allows for dynamical boundary between downlink and uplink subframes, a feature extremely suitable for voice streams with SAD, we shall investigate the performance merit of using dynamical boundary for voice streams with SAD. Our performance measures include packet dropping probability, access delay and channel throughput. In addition, as VoIP supports different codes, we also study the system performance of voice streams with different coding rates.
|Appears in Collections:||[光電與通訊學系] 科技部研究計畫|
Files in This Item:
All items in ASIAIR are protected by copyright, with all rights reserved.