Title: 具聽覺強性的適應性向量預測編碼系統
Vector Adaptive Predictive Coder with Perceptual Enhancements
Authors: 蕭國川
Gwo-Chuan Shiau
張文輝
Dr. Wen-Whei Chang
電信工程研究所
Keywords: 適應性向量預測編碼系統;向量量化;雜訊整形;聽覺強性;;Vector-APC;Vector Quantization;Noise Shaping; Perceptual Enhancement
Issue Date: 1992
Abstract: 由於適應性向量預估編碼系統 (Vector-APC), 引進了向量量化 (VQ) 的
觀念和技術, 這使得它能完成比傳統以純量形式的適應性預估編碼系統
(APC) 更好的語音壓縮品質. 在適應性向量預估編碼系統中, 我們使用長
期純量預估器 (long-delay scalar predictor) 和短期向量預估器
(short-delay vector predictor) 來分別去除語音信號中的長期和短期
冗餘特性, 而後再使用一個適應性增益向量量化器 (gain-adaptive VQ)
去量化所得的預估殘餘信號 (prediction residual). 在這報告中我們提
出了一個位元傳輸率為 9.6 kbits/sec 的加強型適應性向量預估編碼系
統. 這加強型的語音編碼系統提供了比傳統的適應性向量預估編碼系統更
好的聽覺品質. 這是由於我們採用了一個有效的適應性演算法去求得短期
向量預估器的係數和經由一個可靠的濁音/清音判定來分別對濁音和清音
進行編碼. 最後在向量空間發展的雜訊整形濾波器 (noise shaping
fliter) 對於得到一個聽不見的雜訊頻譜也提供相當的幫助.
Vector Adaptive Predictive Coder (VAPC) is a speech coding
algorithm which greatly improves the performance of
conventional scalar Adaptive Predictor Coder (APC) by using
Vector Quantization (VQ) techniques. In VAPC, we use both the
long-delay scalar predictor and the short-delay vector
predictor to remove the long-term and short-term redundancy in
speech waveform respectively, and then apply a gain-adaptive
vector quantizer to quantize the prediction residual. An
enhanced VAPC algorithm for speech coding at 9.6 kbits/s is
presented. It yields better perceptual quality than traditional
VAPC by an efficient adaptive algorithm for solving the short-
delay vector predictor coefficients and separate encoding after
the reliable V-UV decision. Furthermore, the noise shaping
filter developed in vector space also helps to results in an
inaudible noise spectrum.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT810436034
http://hdl.handle.net/11536/57018
Appears in Collections:Thesis