Voice over IP ( VoIP )是ITU-T為了能讓通話雙方透過網際網路交換語音訊息而訂定的一個共通標準，並架構在H.323下。藉由VoIP只需花費少許費用便可以擁有傳統電話網路 ( PSTN) 的功用。 VoIP透過網路傳送，會遭遇到封包延遲的問題是無法避免的，而為了不讓語音封包被網路延遲所影響，我們分析語音的特性及網路的參數，並提出Late-rate動態緩衝區調整及Max-jitter動態緩衝區調整等方式，使的語音在遭受網路延遲時仍能夠有較佳的語音品質;此外，緩衝區的存在會增加語音等待的時間，所以為了減少這種情形，而提出以語音再同步的方式，使的語音在網路延遲，增加緩衝區不會讓語音的播放長度變長，還能有較佳的語音品質。經過實驗的驗證，在有動態緩衝區與再同步的機制下，語音除了能縮短緩衝區時間延遲之外，還能有良好的語音播放品質。 VoIP becomes a very common technique for delivering voice over Internet. It not only costs less but also supports additional services. Due to the lack of QoS support of Internet, the voice quality of VoIP may not be guaranteed. To deal with the packet delay and jitter, we propose two dynamic buffer smoothing techniques, Late-rate method and Max-jitter method, to reduce the damage from delay and jitter. Buffer smoothing can decrease packet loss and yield better quality of speech. However, it also increase the end-to-end delay. Dynamic buffer smoothing is therefore chosen to reduce the average end-to-end delay. In this work, we also propose a resynchronization scheme based on speech silence to adjust the playback length to be normal. Simulations show that the buffer smoothing and the resynchronization techniques really improve the quality of speech significantly.
