AMR-WB语音编码算法的优化及实现
|
摘要
高速发展的DSP技术为语音信号处理提供了强有力的工具,使实时实现各种复杂算法成为可能。针对不同应用,国际标准组织制定了一系列语音压缩编码标准,其中AMR-WB语音编码是3GPP组织于2001年3月制定的自适应变速率宽带语音编码标准,并被国际电信联盟电信标准部ITU-T选定为有线通信的宽带编码标准,其编码方法采用的是代数码本激励线性预测技术。
与其他语音编码标准相比,AMR-WB语音编码器的重建语音具有更高的自然度和可懂度,可广泛应用于移动通信、ISDN宽带电话、ISDN可视电话和电视会议等领域,但其较高的计算复杂度限制了其在计算资源有限的某些领域的应用。本文的目标是在维持合成语音质量的前提下降低AMR-WB语音编码算法计算复杂度,并在TMS320C6416 DSP上实现AMR-WB语音编码的实时处理。
本文对AMR-WB语音编码算法的时间复杂度、空间复杂度及合成语音质量进行了评估。评估测试表明:固定码本搜索算法的计算复杂度为674.13MIPS,占整个AMR-WB语音编码算法的40%左右。这是导致整个语音编码算法复杂度偏高的主要原因。针对上述问题,本文在充分研究了ACELP固定码本搜索方法的基础上,提出了一种基于AMR-WB固定码本结构的快速搜索方法,实验结果表明,与AMR-WB中深度优先树搜索方法相比,新方法不仅能够维持合成语音质量,更使固定码本搜索模块的计算复杂度降低了53.6%,整个语音编码算法的计算复杂度降低了21.2%;最后,对AMR-WB语音编码算法进行了基于TMS320C6416平台的移植和优化,完成了关键模块的C代码优化,实时实现了AMR-WB语音编码算法。
With the rapid development of DSP technology, it becomes a powerful tool in the speech signal processing, which makes it possible to implement all kinds of complex algorithms in real time. Based on different applications, many standards of speech compress coding have been established by different international standardization organizations. The AMR-WB speech codec based on ACELP was approved by 3GPP in March 2001 and selected for wideband speech coding by ITU-T.
Compared with other speech coding standards, AMR-WB is much more natural and intelligible, so it is widely used in Mobile Communications, ISDN wideband telephone, ISDN video telephone and video conferencing. But its high complexity limits its application in the area with critical hardware platform. The purpose of the thesis is to lower the complexity of AMR-WB, ensure the coding quality and implement the AMR-WB speech coding algorithm in real time on the platform of TMS320C6416.
This thesis evaluated the computing complexity, memory space complexity and coding quality of AMR-WB coding algorithm. The results show that the computing complexity of fixed codebook search algorithm is 674.13MIPS, occupying 40% of the whole AMR-WB coding algorithm. This results in the high computing complexity of AMR-WB coding algorithm. In order to reducing the computing complexity of AMR-WB coding algorithm, this thesis proposed a fast fixed codebook search algorithm for AMR-WB coding algorithm on the basis of studying the methods of ACELP fixed codebook search. The results show that the computing complexity of the new algorithm reduces by 53.6% compared with deep first tree search algorithm in AMR-WB,not only reduces the computing complexity of the whole coding algorithm significantly, but also ensures the coding quality. Lastly, this thesis transported and optimized the AMR-WB coding algorithm with improved algorithm based on the platform of TMS320C6416 DSP, completed the optimization of kernal algorithms at the C code level and real-time implementation of the AMR-WB speech coding algorithm.
与其他语音编码标准相比,AMR-WB语音编码器的重建语音具有更高的自然度和可懂度,可广泛应用于移动通信、ISDN宽带电话、ISDN可视电话和电视会议等领域,但其较高的计算复杂度限制了其在计算资源有限的某些领域的应用。本文的目标是在维持合成语音质量的前提下降低AMR-WB语音编码算法计算复杂度,并在TMS320C6416 DSP上实现AMR-WB语音编码的实时处理。
本文对AMR-WB语音编码算法的时间复杂度、空间复杂度及合成语音质量进行了评估。评估测试表明:固定码本搜索算法的计算复杂度为674.13MIPS,占整个AMR-WB语音编码算法的40%左右。这是导致整个语音编码算法复杂度偏高的主要原因。针对上述问题,本文在充分研究了ACELP固定码本搜索方法的基础上,提出了一种基于AMR-WB固定码本结构的快速搜索方法,实验结果表明,与AMR-WB中深度优先树搜索方法相比,新方法不仅能够维持合成语音质量,更使固定码本搜索模块的计算复杂度降低了53.6%,整个语音编码算法的计算复杂度降低了21.2%;最后,对AMR-WB语音编码算法进行了基于TMS320C6416平台的移植和优化,完成了关键模块的C代码优化,实时实现了AMR-WB语音编码算法。
With the rapid development of DSP technology, it becomes a powerful tool in the speech signal processing, which makes it possible to implement all kinds of complex algorithms in real time. Based on different applications, many standards of speech compress coding have been established by different international standardization organizations. The AMR-WB speech codec based on ACELP was approved by 3GPP in March 2001 and selected for wideband speech coding by ITU-T.
Compared with other speech coding standards, AMR-WB is much more natural and intelligible, so it is widely used in Mobile Communications, ISDN wideband telephone, ISDN video telephone and video conferencing. But its high complexity limits its application in the area with critical hardware platform. The purpose of the thesis is to lower the complexity of AMR-WB, ensure the coding quality and implement the AMR-WB speech coding algorithm in real time on the platform of TMS320C6416.
This thesis evaluated the computing complexity, memory space complexity and coding quality of AMR-WB coding algorithm. The results show that the computing complexity of fixed codebook search algorithm is 674.13MIPS, occupying 40% of the whole AMR-WB coding algorithm. This results in the high computing complexity of AMR-WB coding algorithm. In order to reducing the computing complexity of AMR-WB coding algorithm, this thesis proposed a fast fixed codebook search algorithm for AMR-WB coding algorithm on the basis of studying the methods of ACELP fixed codebook search. The results show that the computing complexity of the new algorithm reduces by 53.6% compared with deep first tree search algorithm in AMR-WB,not only reduces the computing complexity of the whole coding algorithm significantly, but also ensures the coding quality. Lastly, this thesis transported and optimized the AMR-WB coding algorithm with improved algorithm based on the platform of TMS320C6416 DSP, completed the optimization of kernal algorithms at the C code level and real-time implementation of the AMR-WB speech coding algorithm.
引文
[1]王炳锡.变速率语音编码[M].西安:西安电子科技大学出版社, 2004: 1-97.
[2] TIA/EIA/96-C. Speech Service Option Standard for Wideband Spread Spectrum Systems[S]. Washington: Telecommunications Industry Association, 1998.
[3] TIA/EIA/IS-733. High Rate Speech Service Option 17 for Wideband Spread Sprectrum Communication Syetems[S]. Washington: Telecommunications Industry Association, 1998.
[4] TS26.071 V3.0.1. AMR Speech Codec about General Description[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications , 1999.
[5] TDOC SMG11 170/991. On Wideband Codec performance[S]. Finland: Nokia ETSI , 1999 .
[6] VoiceAge. Wideband Speech Coding Standards and Applications[S]. Montreal : VoiceAge, 2001.
[7] Varga I, Drogo R, Iacovo D, etal. Standardization of the AMR Wideband Speech Codec in 3GPP and ITU-T[J]. IEEE Communications Magazine, 2006, 44(5): 66-73.
[8]王俊翔. GSM新技术及应用前景展望[J].电信网技术, 2007, 6(): 44-48.
[9] SPRA656. Implementation of G.729 on the TMS320C54x[S]. Dallas: Texas Instruments Incorporated, 2000.
[10] Zhu Z B, Wang Q. Speech Codec Optimization Based On Cell Broadband Engine[A]. In: ICASSP’2007[C]. Hawaii: IEEE Press, 2007: 805-808.
[11] Jee D G, Hoi S I. Real-time Implement of AMR-WB Speech Coder Using TMS32OC5509 DSP[A]. In: ICACT’2005[C]. Korea: IEEE Press, 2005: 1387-1390.
[12] Byun K J, Eo I S, Jeong H B. Real-time Implementation of AMR and AMR-WB Using the Fixed-point DSP for WCDMA Systems[A]. In: ISCE’2006[C]. St.Petersbuig: IEEE Press, 2006: 153-159.
[13] TS26.171 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec: General Desciription[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2006.
[14] TS26.193 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec:Sourcecontrolled rate operation[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2007.
[15] TS26.191 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec:error concealment of erroneous or lost frame[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2007.
[16] TS26.190. Adaptive Multi-Rate WideBand Speech Transcoding[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2000.
[17] Makhoul J. Linear Prediction: A Tutorial Review[J]. Proceeding of the IEEE, 1975, 63(4): 561-580.
[18] TS26.194 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec: Voice Activity Detector[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2007.
[19] TS26.192 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec: Comfort noise aspects[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2007.
[20] G.722.2 Annex D: Digital test sequence[S]. Switzerland : ITU-T, 2002.
[21]李方慧,王飞,何佩坤. TMS320C6000系列DSPs原理与应用[M].北京:电子工业出版社, 2005: 110-134.
[22] SPRU395B. TMS320C64x Technical Overview[S]. Dallas: Texas Instruments Incorporated ,2001.
[23] Recommendation P.862: Perceptual evaluation of speech quality (PESQ), an objective method for end-to-end speech quality assessment of narrowband telephone networks and speech codecs[S]. Switzerland : ITU-T, 2002.
[24] P.862 Ammendment1.Source code for reference implementation and conformance tests[S]. Switzerland : ITU-T, 2002.
[25]鲍长春.数字语音编码原理[M].西安:西安电子科技大学出版社, 2007: 171-189.
[26] Byun K J, Jung H B, Hahh M, etal. A fast ACELP Codebook Search Method[A]. In: Signal Processing’2002[C]. BangKok: IEEE Press, 2002: 422-425.
[27] Park H, Choi Y, Lee D. efficient codebook search method for acelp speech codecs[A]. In: Speech Coding’2002[C]. Lbaraki: IEEE Press, 2002: 17-19.
[28] Lee E D, Lee M S, Kim D Y. Global Pulse Replacement Method for Fixed Codebook Search of ACELP Speech Codec[A]. In: The 2nd IASTED International Conference on Communication, Internet, and Information Technology[C]. Scottsdale: IEEE Press, 2003: 372-375.
[29] Lee E D, Ahn J M. Efficient Fixed Codebook Search Method for ACELP Speech Codecs[A]. In: ICHIT 2006 [C]. Jeju Island: Springer-Verlag Berlin Heidelberg, 2007: 178-187.
[30] Lee E D, Yun S H, Lee S I, etal. Iteration-free pulse replacement method for algebraic codebook search[J]. ELECTRONICS LETTERS, 2007, 43(1): 59-60.
[31]朱敏,朱小康.一个快速的码本搜索方法[J].通信技术. 2007, 7(): 73-75.
[32] Ha N K. A fast search method of algebraic codebook by reordering search sequence[A]. In: ICASSP’99[C]. Phoenix: IEEE Press, 1999: 21-24.
[33] Chen F K, Yang J F. Maximum take precedence ACELP: a low complexity search method[A]. In: ICASSP’01[C]. Salt Lake City: IEEE Press, 2001: 693-696.
[34] PravinKumar R. High computational performance in code exited linear prediction speech model using faster codebook search techniques[A]. In: International Conference on Computing:Theory and Applications, ICCTA’07[C]. Kolkata: IEEE Press, 2007: 458-462.
[35] Yu S S, Yang J, Zhou J L. Algorithm improving the CELP coder for real-time communication[A]. In: IEEE proceeding of the 2008 Congress on Image and Signal Processing[C] . Hainan: IEEE Press, 2008: 565-567.
[36] Falahati A, Soleimani M, Tabataba Vakili V. Dynamic tree pruning method for fast ACELP search[A]. In: Information and Communication Thechnologies: From Theory to Applications, ICTTA 2008[C]. Damascus: IEEE Press, 2008: 1-6.
[37] Wang M L, Yang J F. Generalised candidate scheme for the stochastic codebook search of scalable CELP coders[J]. Vision, Image and Signal Processing, 2004, 151(5): 443-452.
[38] Salami R, Laflamme C, Bessette B, etal. ITU-T G.729 Aneex A: reduced complexity 8kb/s CS-ACELP codec for Digital simultaneous voice and data[J]. Communications Magazine, 1997, 35(9): 56-63.
[39] Falahati A, Soleimani M. A proposed fast ACELP codebook search[A]. In: APCC’2008[C]. Tokyo: IEEE Press, 2008: 1-5.
[40] SPRU610. TMS320C64xDSP Two-Level Internal Memory Reference Guide[S]. Dallas: Texas Instruments Incorporated, 2004.
[41] SPRU189F. TMS320C6000 CPU and Instruction set Reference Guide[S]. Dallas: Texas Instruments Incorporated , 2000.
[42] SPRU301C. TMS320C6000 Code Composer Studio User’s Guide[S]. Dallas:Texas Instruments Incorporated ,2000.
[43]阳明晔,张志勇.基于TMS320C6000系列DSP的C代码优化方法研究[J].微处理机, 2004, (2): 59-64.
[44]林峰,林毅. TMS320C6000代码优化技术[J].重庆邮电学院学报(自然科学版), 2006, 18(1): 60-64
[45] SPRU187I. TMS320C6000 Optimizing C/C++ Complier User’s Guide[S]. Dallas: Texas Instruments Incorporated , 2001.
[46] SPRS146N. TMS320C6416, TMS320C6415 Fixed-Point Digital Signal Processors Data Manual[S]. Dallas: Texas Instruments Incorporated , 2005.
[47] Lee E D, Kim D Y, Kim B T. Focused search Method of fixed codebook and Apparatus thereof [P]. United States Patent: US7302386B2, 2007-11-27.
[48] Lee D R. Excitation codebook search method in a speech coding system[P]. United States Patent: US2007/0043560 A1, 2007-2-22.
[2] TIA/EIA/96-C. Speech Service Option Standard for Wideband Spread Spectrum Systems[S]. Washington: Telecommunications Industry Association, 1998.
[3] TIA/EIA/IS-733. High Rate Speech Service Option 17 for Wideband Spread Sprectrum Communication Syetems[S]. Washington: Telecommunications Industry Association, 1998.
[4] TS26.071 V3.0.1. AMR Speech Codec about General Description[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications , 1999.
[5] TDOC SMG11 170/991. On Wideband Codec performance[S]. Finland: Nokia ETSI , 1999 .
[6] VoiceAge. Wideband Speech Coding Standards and Applications[S]. Montreal : VoiceAge, 2001.
[7] Varga I, Drogo R, Iacovo D, etal. Standardization of the AMR Wideband Speech Codec in 3GPP and ITU-T[J]. IEEE Communications Magazine, 2006, 44(5): 66-73.
[8]王俊翔. GSM新技术及应用前景展望[J].电信网技术, 2007, 6(): 44-48.
[9] SPRA656. Implementation of G.729 on the TMS320C54x[S]. Dallas: Texas Instruments Incorporated, 2000.
[10] Zhu Z B, Wang Q. Speech Codec Optimization Based On Cell Broadband Engine[A]. In: ICASSP’2007[C]. Hawaii: IEEE Press, 2007: 805-808.
[11] Jee D G, Hoi S I. Real-time Implement of AMR-WB Speech Coder Using TMS32OC5509 DSP[A]. In: ICACT’2005[C]. Korea: IEEE Press, 2005: 1387-1390.
[12] Byun K J, Eo I S, Jeong H B. Real-time Implementation of AMR and AMR-WB Using the Fixed-point DSP for WCDMA Systems[A]. In: ISCE’2006[C]. St.Petersbuig: IEEE Press, 2006: 153-159.
[13] TS26.171 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec: General Desciription[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2006.
[14] TS26.193 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec:Sourcecontrolled rate operation[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2007.
[15] TS26.191 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec:error concealment of erroneous or lost frame[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2007.
[16] TS26.190. Adaptive Multi-Rate WideBand Speech Transcoding[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2000.
[17] Makhoul J. Linear Prediction: A Tutorial Review[J]. Proceeding of the IEEE, 1975, 63(4): 561-580.
[18] TS26.194 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec: Voice Activity Detector[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2007.
[19] TS26.192 V7.0.0. Adaptive Multi-Rate WideBand Speech Codec: Comfort noise aspects[S]. Sophia Antipolis Valbonne: 3GPP Organisational Partners' Publications, 2007.
[20] G.722.2 Annex D: Digital test sequence[S]. Switzerland : ITU-T, 2002.
[21]李方慧,王飞,何佩坤. TMS320C6000系列DSPs原理与应用[M].北京:电子工业出版社, 2005: 110-134.
[22] SPRU395B. TMS320C64x Technical Overview[S]. Dallas: Texas Instruments Incorporated ,2001.
[23] Recommendation P.862: Perceptual evaluation of speech quality (PESQ), an objective method for end-to-end speech quality assessment of narrowband telephone networks and speech codecs[S]. Switzerland : ITU-T, 2002.
[24] P.862 Ammendment1.Source code for reference implementation and conformance tests[S]. Switzerland : ITU-T, 2002.
[25]鲍长春.数字语音编码原理[M].西安:西安电子科技大学出版社, 2007: 171-189.
[26] Byun K J, Jung H B, Hahh M, etal. A fast ACELP Codebook Search Method[A]. In: Signal Processing’2002[C]. BangKok: IEEE Press, 2002: 422-425.
[27] Park H, Choi Y, Lee D. efficient codebook search method for acelp speech codecs[A]. In: Speech Coding’2002[C]. Lbaraki: IEEE Press, 2002: 17-19.
[28] Lee E D, Lee M S, Kim D Y. Global Pulse Replacement Method for Fixed Codebook Search of ACELP Speech Codec[A]. In: The 2nd IASTED International Conference on Communication, Internet, and Information Technology[C]. Scottsdale: IEEE Press, 2003: 372-375.
[29] Lee E D, Ahn J M. Efficient Fixed Codebook Search Method for ACELP Speech Codecs[A]. In: ICHIT 2006 [C]. Jeju Island: Springer-Verlag Berlin Heidelberg, 2007: 178-187.
[30] Lee E D, Yun S H, Lee S I, etal. Iteration-free pulse replacement method for algebraic codebook search[J]. ELECTRONICS LETTERS, 2007, 43(1): 59-60.
[31]朱敏,朱小康.一个快速的码本搜索方法[J].通信技术. 2007, 7(): 73-75.
[32] Ha N K. A fast search method of algebraic codebook by reordering search sequence[A]. In: ICASSP’99[C]. Phoenix: IEEE Press, 1999: 21-24.
[33] Chen F K, Yang J F. Maximum take precedence ACELP: a low complexity search method[A]. In: ICASSP’01[C]. Salt Lake City: IEEE Press, 2001: 693-696.
[34] PravinKumar R. High computational performance in code exited linear prediction speech model using faster codebook search techniques[A]. In: International Conference on Computing:Theory and Applications, ICCTA’07[C]. Kolkata: IEEE Press, 2007: 458-462.
[35] Yu S S, Yang J, Zhou J L. Algorithm improving the CELP coder for real-time communication[A]. In: IEEE proceeding of the 2008 Congress on Image and Signal Processing[C] . Hainan: IEEE Press, 2008: 565-567.
[36] Falahati A, Soleimani M, Tabataba Vakili V. Dynamic tree pruning method for fast ACELP search[A]. In: Information and Communication Thechnologies: From Theory to Applications, ICTTA 2008[C]. Damascus: IEEE Press, 2008: 1-6.
[37] Wang M L, Yang J F. Generalised candidate scheme for the stochastic codebook search of scalable CELP coders[J]. Vision, Image and Signal Processing, 2004, 151(5): 443-452.
[38] Salami R, Laflamme C, Bessette B, etal. ITU-T G.729 Aneex A: reduced complexity 8kb/s CS-ACELP codec for Digital simultaneous voice and data[J]. Communications Magazine, 1997, 35(9): 56-63.
[39] Falahati A, Soleimani M. A proposed fast ACELP codebook search[A]. In: APCC’2008[C]. Tokyo: IEEE Press, 2008: 1-5.
[40] SPRU610. TMS320C64xDSP Two-Level Internal Memory Reference Guide[S]. Dallas: Texas Instruments Incorporated, 2004.
[41] SPRU189F. TMS320C6000 CPU and Instruction set Reference Guide[S]. Dallas: Texas Instruments Incorporated , 2000.
[42] SPRU301C. TMS320C6000 Code Composer Studio User’s Guide[S]. Dallas:Texas Instruments Incorporated ,2000.
[43]阳明晔,张志勇.基于TMS320C6000系列DSP的C代码优化方法研究[J].微处理机, 2004, (2): 59-64.
[44]林峰,林毅. TMS320C6000代码优化技术[J].重庆邮电学院学报(自然科学版), 2006, 18(1): 60-64
[45] SPRU187I. TMS320C6000 Optimizing C/C++ Complier User’s Guide[S]. Dallas: Texas Instruments Incorporated , 2001.
[46] SPRS146N. TMS320C6416, TMS320C6415 Fixed-Point Digital Signal Processors Data Manual[S]. Dallas: Texas Instruments Incorporated , 2005.
[47] Lee E D, Kim D Y, Kim B T. Focused search Method of fixed codebook and Apparatus thereof [P]. United States Patent: US7302386B2, 2007-11-27.
[48] Lee D R. Excitation codebook search method in a speech coding system[P]. United States Patent: US2007/0043560 A1, 2007-2-22.