增量調變語音封包之漏失回覆技術

Abstract

本文主要是探討在分封式網路中，如何針對封包漏失的問題進行訊號的 回覆與修補,首先 探討一種包覆式編碼(embedded coding)架構，依低階 位元棄置(LSB-dorpping)的原則處 理，但此方法並不適用在具有嚴重封 包漏失的通訊環境中。在奇偶分封的 傳輸模式下，取樣插補是最常見的 回覆架構。它是利用前後相鄰兩點的線性插補值 做為漏失點的近似值。 然而；針對本文所使用的適應性增量調變編碼系統而言，取 樣插補並無 法有效解決封包漏失對其量化步階所造成的脫序現象。因此，本文將提 出以插補式增量調變為基礎的樹狀插補模式，利用樹狀插補的多路徑搜尋 原理，加 強系統在封包漏失的回覆能力。 Packet communication increases the utilization efficiency of channel capacity by exploiting the bursty nature of voice traffic. Depending upon the contents of speech packets, numerous recovery techniques have been proposed to combat the packet loss due to channel impairments or excessive delay. Most previous research has emphasized the pissing packet recovery techniques for PCM and DPCM coded speech. When using a DM encoder, however, the missing samples in the discarded packets cannot be reconstructed by zero-amplitude stuffing or waveform substitution techniques. In this thesis, we show that the packet loss effects of DM system or using a tree search interpolation. In this work, we first provide experimental results for EDM coding of speech sources under packet loss conditions. For the tree search interpolation, we explore the benefits of delayed decoding by using an interpolative DM code generator to form a tree of sample possibilities given the remaining adjacent samples.