本論文實作分散式多人網路語音會議系統,廣泛應用於即時網路語音會議。本系統分為伺服端與用戶端兩個部份。伺服端的部份,包含兩種伺服器,分別為認證伺服器以及通話管理伺服器。認證伺服器主要是進行用戶的授權與認證;而通話管理伺服器則是負責語音連線的管理。在伺服端的實作開發上,本論文採用反應器(Reactor)設計模式來實現。反應器具有獨立的事件解析以及事件處理器分派能力,不僅能提升伺服器在處理網路事件上的效能與穩定性,更能增加開發速度及程式品質。 用戶端的部份,本論文提出多人語音混音機制與分散式會議控制協定。利用多人語音混音,不但能讓用戶進行多方會談,體驗更接近真實的語音通話,還能避免在即按即說(Push To Talk)模式下,用戶獨佔語音資源的情況發生;分散式會議控制協定則是負責會議連線的管理,包含會議成員的加入與離開、用戶端與伺服端之間的訊息交換以及會議多媒體連線的建立。 本論文所設計的系統,經由實作測試,證實在多人同時連線的情況下,仍然能保持穩定的反應時間,同時提供清晰流暢的語音品質。
We have implemented the distributed multiparty conferencing system for voice over internet protocol that applies to real-time voice over IP conferencing extensively. It divides into a server part and a client part. The server part comprises one AAA (Authentication Authorization Accounting) server for user’s authentication and authorization and one CM (Call Management) server for manage voice calls. In our server, we adopt the Reactor design pattern to implement. Reactor provides capability for event driven and event handler dispatch, not only improves the performance and stability for server to process network event, but also enhances the development speed and quality of code. In client part, we propose multiparty speech mixing and distributed conferencing control protocol. Apply multiparty speech mixing, not only can make user talking to each other at the same time, moreover experience more real speech environment on the internet, but also can avoid the speech resource monopolized bye push to talk mode. On the other hand, the distributed conferencing control protocol is responsible for manage the conferencing call, include conferencing member’s joining and leaving, information exchange between client and server and set up the conferencing multimedia connection. Our system through real testing in network environment, verify that can still keep steady response time under multiple connections, furthermore provide clear and smooth speech quality.