臺灣博碩士論文加值系統

本論文主要研究內容是針對MPEG-4 中低位元率語音編碼標準−諧波向量激發編碼器(Harmonic Vector eXcitation Coder，簡稱HVXC )，做演算法之快速實現。首先，在編碼端提出一種多階段有聲/無聲決定的方法，使得整個編碼端能隨著語音信號特性的不同，在不降低原來合成語音品質的條件下，適應性地改變整個演算架構的流程，以節省不必要的計算量。接下來，因求取諧波振幅過程中，必先求得精確的基週，我們提出一種樹狀搜尋法，而且能隨著語音信號本身特性不同，適應性地改變搜尋的範圍，有效率地節省一些不必要的計算量。

In this thesis, we develop fast algorithms for the MPEG-4 low bit-rateHVXC speech coder to adaptively reduce the computation. First, we proposea novel multi-stage voiced/unvoiced decision algorithm. According to thecharacteristics of the encoding speech signal, each detected stage adaptively
adjusts the flow of the encoder to reduce the unnecessary computation and achieves the same speech quality. In the HVXC speech coder, the pitch is an indispensable parameter for successful estimation of the harmonic magnitudes.
Hence, we secondly propose a tree searching method to refine the estimated pitch. The propose method can adaptively amend the search range of candidate pitches such that we can further reduce the needless computation in accordance with the characteristics of the processing speech signal.

大綱目錄 ···············I
表格目錄 ···············V
圖示目錄 ·············· VI
第一章簡介 ··············1
1.1 背景及動機 ············1
1.2 論文大要 ·············3
第二章HVXC 編碼端主要演算法 ····· 5
2.1 簡介 ···············5
2.2 線性預測之分析 ··········7
2.3 有聲語音部分的演算法 ·······9
2.3.1 頻譜封包向量之估測 ·······9
2.3.2 有聲及無聲的頻帶的判別·····11
2.3.3 維度的轉換···········14
2.4 無聲語音部分的演算法·······18
2.4.1 碼激式線性預測編碼器(CELP)架構·19
2.4.2 知覺加權濾波器·········22
2.4.3 知覺加權LPC 合成濾波器·····22
2.4.4 最佳碼激向量搜尋演算法·····23
2.4.5 使用“向後濾波器”的快速碼簿搜尋法··25
第三章HVXC解碼端主要演算法······30
3.1 參數的調整············30
3.1.1 基週參數的調整·········31
3.1.2 時間參數的內插·········31
3.2 有聲區間語音合成·········35
3.2.1 有聲區間諧波激發信號的產生···36
3.2.2 諧波激發信號的產生·······39
3.2.3 有聲區間LPC 合成濾波器·····45
3.3 無聲區間語音合成·········46
3.3.1 無聲區間激發信號的產生·····46
3.3.2 無聲有聲激發信號轉變的平滑化··47
3.3.3 無聲區間LPC 合成濾波器·····48
3.4 後置濾波器············50
3.4.1 有聲語音的後置濾波器······50
3.4.2 無聲語音的後置濾波器······52
3.4.3 後置處理············53
第四章有聲/無聲判定結構的改進···· 55
4.1 簡介···············55
4.2 標準HVXC 有聲/無聲(V/UV)的決定··56
4.3 多階段有聲/無聲(V/UV)決定的機制·60
4.3.1 簡介··············60
4.3.2 第一階段有聲/無聲的決定 ····61
4.3.3 第二階段有聲/無聲的決定 ····67
4.3.4 第三階段有聲/無聲的決定 ····68
4.4 訓練及測試結果··········69
4.4.1 測試設計············69
4.4.2 訓練的結果···········70
4.4.3 測試的結果···········72
4.4.4 V/UV 錯誤音框分析及討論 ····73
4.5 效能分析·············74
第五章求取諧波振幅的改進方法·····78
5.1 簡介···············78
5.2 HVXC標準求取諧波振幅的方法····79
5.2.1 求取諧波振幅的步驟·······79
5.2.2 求取諧波振幅的方法·······79
5.3 求取諧波振幅的改進方法······81
5.3.1 求取諧波振幅的改進方法·····81
5.3.2 效能分析············83
5.4 有效率地求取精確基週·······84
5.4.1 有效率地求取精確基週改進方法··84
5.4.2 確定搜尋範圍的特徵·······86
5.4.3 測試結果············88
5.4.4 效能分析············90
第六章結論··············91
參考文獻···············93

[1] L. R. Rabiner and R. W. Schafer, “Digital Processing of Speech Signals”, 1978
[2] A. M. Kondoz, “Digital Speech-Coding for Low Rate Communications Systems”,April 1999
[3] D. W. Griffin and J. S. Lim, “Multiband Excitation Vocoder”, IEEE Trans. on ASSP,664-678, August 1988
[4] P. C. Meuse, “A 2400 bps Muti-Band Excitation Vocoder”, Proc. ICASSP, Page(s):9 —12, 1990
[5] Nishiguchi, M.; Matsumoto, J, “Vector quantized MBE with simplified V/UV division at 3.0 kbit/s”, ICASSP, Page(s): 151 -154 vol.2, 1993
[6] R. J. McAulay and T. F. Quatieri, “Speech Analysis/Synthesis Based on a Sinusoidal Representation”, IEEE Trans. ASSP, Vol.34, No 4, pp.744-754, Aug 1986
[7] Y. Shoham, “High-Quality Speech Coding at 2.4 to 4.0 Kbps based on Time-Frequency Interpolation”, Proc. ICASSP, pp.II-151-154, Apr. 1993
[8] M. Nishiguchi; A. Inoue; Y. Maeda and J. Matsumoto, “Parametric Speech Coding-HVXC at 2.0-4.0kbps”, Speech Coding Proceedings, IEEE Workshop, Page(s):84—86, 1999
[9] M. Nishiguchi and J. Matsumoto, “Harmonic and noise coding of LPC residuals with classified vector quantization”, ICASSP, Volume: 1, Page(s): 484 -487 vol.1, 1995
[10] ISO/IEC JTC 1/SC 29/WG 11 N2503-2H, 1998-11-15, “Information technology—Coding of audio-visual objects, Part 3: Audio, Subpart 2: Speech Coding - HVXC.
[11] 連桂宏, MPEG-4 低率語音編碼器-HVXC 編碼器之實現, 碩士論文--國立成功大學電機工程研究所, 民90