標題: | 第三代無線通訊語音編碼之模擬與實現 Simulation and Implementation of AMR Speech Coding for 3GPP |
作者: | 陳景川 Jing-Chuan Chen 杭學鳴 Hsueh-Ming Hang 電子研究所 |
關鍵字: | 語音編碼;無線通訊;語音品質;數位訊號處理器;AMR;3GPP;DSP;speech quality;G.723.1;SNR;spectral dsitance |
公開日期: | 2000 |
摘要: | 第 三 代 無 線 通 訊
語 音 編 碼 之 模 擬 與 實 現
研究生: 陳景川 指導教授: 杭學鳴博士
國立交通大學
電子工程學系 電子研究所碩士班
摘要
一直以來,語音服務都是無線通訊中的核心部分,它在下一世代的無線通訊系統中也將持續扮演重要的角色。本篇論文將焦點放在一個稱為「適應性多重位元率」( AMR)的語音編碼標準。它是由「第三代合作計畫」(3GPP)這個組織所提出的。這個組織的目標在於推動並訂立下一代無線通訊系統相關的標準和規格。而AMR是這一組規格中不可或缺的要素。AMR語音壓縮標準是以「電碼激發線性估計編碼」(CELP)的技術為設計基礎,並且有八種編碼率,分別介於4.75到12.2 kbit/s之間。
在本篇論文中,我們首先回顧此一壓縮標準的簡史,討論並以公式明確地陳述此標準所含括的概念和原則。接著,我們介紹評估語音品質的方法,並提出兩種客觀的語音品質度量法則,分別為segmental SNR及spectral distance。除此之外,為了將無線傳輸通道中的噪音以及漸強漸弱現象對語音品質所造成的影響考慮進來,我們引進兩種通道雜訊的模型,分別是「加成性白色高斯雜訊」(AWGN)和「雷利衰減通道」(Rayleigh fading channel)。我們針對「國際電信聯盟」(ITU)的G.723.1以及3GPP的AMR這兩種壓縮標準,比較其語音品質之高下。實驗數據顯示,無論是在無干擾的通道或是在有雜訊影響的情形之下,「適應性多重位元率」都有較好的語音品質。若以spectral distance來評估,在無干擾的通道中,AMR領先G.723.1將近1 dB;在位元錯誤率(bit error rate)達到5x10-4時,AMR仍以0.6dB到0.9dB的幅度領先G.723.1。
除此之外,我們利用德州儀器工業公司(TI)的數位訊號處理器來實現「適應性多重位元率」語音壓縮標準。為了加速編解碼器執行的速率並且縮小程式碼所佔的空間,我們對程式碼做修改和最佳化。如此一來,編解碼器就能充分利用這個數位訊號處理器所提供的硬體資源。藉由適當的修改和最佳化之後,和原來的情形相較,這個「適應性多重位元率」編解碼器節省了百分之四十的執行時間,而程式碼所佔的空間也削減為原來的百分之七十。 Simulation and Implementation of AMR Speech Coding for 3GPP Student: Jing-Chuan Chen Advisor: Dr. Hsueh-Ming Hang Institute of Electronics Department of Electronics Engineering National Chiao Tung University Abstract Speech service has always been the central part in the wireless communication, and it will still be one of the most important services in the next generation of wireless telecommunication systems. This thesis focuses on a recent speech coding standard, Adaptive Multi Rate (AMR) coding, defined by the Third Generation Partnership Project (3GPP). As the name implies, 3GPP’s goal is to define a set of new specifications for the next generation wireless communication system. The AMR speech coder is an integral part of this set of specifications. AMR is designed based on the code-excited linear predictive (CELP) coding technique, and it can operate at various bit rates between 4.75 and 12.2 kbit/s. After briefly reviewing the history of AMR, we describe the concepts and principles of AMR. Then, two speech quality measures, segmental signal-to-noise ratio (segSNR) and spectral distance (SD), are introduced for speech quality assessment. We simulate and compare the compressed speech quality of 3GPP AMR and ITU-T G.723.1 at various bit rates and channel conditions. Two channel error models, additive white Gaussian noise (AWGN) channel and Rayleigh fading channel, have been used for performance evaluation. The simulation results indicate that AMR exceeds G.723.1 by about 1dB in spectral distance, either under the clean channel or under a noisy channel, which has a bit error rate (BER) about 5x10-4. Moreover, we implement AMR speech codec on the Texas Instrument (TI) TMS320C6701 digital signal processor (DSP). In order to speed up and reduce code size, we modify and optimize the C codes. The optimized AMR codec can fully utilize the resources of the TI DSP chips. With modifications and optimizations, the codec can save 40% of execution time, compared to the original version. In the same time, the code size is reduced down to 70% of the original. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#NT890428062 http://hdl.handle.net/11536/67137 |
顯示於類別: | 畢業論文 |