Abstract

Factors serving as constraints in today’s communication systems include bandwidth consumption, power, quality and delay. The desire to communicate effectively and exchange as much information as possible by transmitting only a few parameters, derives an entire industry. In Voice over Internet Protocol (VoIP) application, that require the transmission of speech, all these goals are addressed by development of efficient methods of speech compression. This paper compares two speech-encoding algorithms, differentiated by their fundamental approach to coding and decoding of speech signals. MPMLQ, Multi Pulse Maximum Likelihood Quantization, that separately encode the even and odd pulse positions by employing combinatorial coding. ACELP, Algebraic Code Excited Linear Prediction, is an enhanced version of CELP, Code Excited Linear Prediction speech coding algorithms, which employ an algebric codebook for efficient searching. These coding techniques will be compared based on the delay, computational load distribution, and no of bits transmitted for each technique, discussing the fundamentals behind each and concluding factors and result important to system designers, implementers and subscribers.

Keywords

MPMLQ, ACELP, ITU, Speech compression, CELP.

Extended Summary

Although with the emergence of optical fibers bandwidth in wired communications has become inexpensive still, there is a growing need for bandwidth conservation and enhanced privacy in wireless cellular and satellite communications. In particular, cellular communications have been enjoying a tremendous worldwide growth and there is a great deal of Research and Development (R&D) activity geared towards establishing global portable communications through wireless personal communication networks. On the other hand, there is a trend toward integrating voice-related applications (e.g., voicemail) on desktop and portable personal computers - often in the context of multimedia communications. Most of these applications require that the speech signal is in digital format so that it can be processed, stored, or transmitted under software control. This paper compares two of these speech coding techniques, based on the means by which they achieve compression: MPMLQ and ACELP. , Multi Pulse Maximum Likelihood Quantization, MPMLQ, and Algebric Code Excited Linear Prediction, ACELP, make an interesting pair because they addresses speech compression using two fundamentally different techniques, and both of these algorithms are part of the International Telecommunication Union (ITU) Recommendation G.723.1.

Speech is generally band limited to 4 kHz (or 3.2 kHz) and sampled at 8kHz, although digital speech brings flexibility and opportunities for encryption, it is also associated (when uncompressed) with a high data rate and hence high requirements of transmission bandwidth and storage. Speech Coding or Speech Compression is the field concerned with obtaining compact digital representations of voice signals for the purpose of efficient transmission or storage. Speech coding involves sampling and amplitude quantization. The objective in speech coding is to represent speech with a minimum number of bits while maintaining its perceptual quality. Speech coders differ widely in their approaches to achieving signal compression. Based on the means by which they achieve compression: MPMLQ and ACELP are classified as analysis-by-synthesis coders. Analysis-by-synthesis speech coders are among the newest and most effective of modern speech coders

The results demonstrate the delay, computational load distribution, and no of bits transmitted for each technique, obtained by testing our codec using the test vectors provided by ITU-T. The objective of this paper is to compare two fundamentally different speech-coding techniques, MPMLQ and ACELP implemented under ITU-T G.723.1 recommendation. We conclude form the results that the delay associated with MPMLQ is slightly greater than that of the ACELP. Next we analyzed the computation load distribution graphs, which shows the complexity level associated with each algorithm. It can be concluded form the graphs that MPMLQ stochastic codebook search takes 55% of the time where as ACELP stochastic codebook search takes 47% of the time. The bit allocation graphs demonstrate that ACELP takes less no of bits to transmit than MPMLQ. All these proven results make ACELP an attractive solution for low delay speech compression applications. However, the major tradeoff exists between the toll quality speech reconstructed using MPMLQ and the lower transmission rate and delay of ACELP technique, providing system designers, implementers and subscribers with flexibility.

Figures

Figure 1. Even and Odd pulse positions for MPMLQ

Figure 2: Delay associated with MPMLQ

Figure 3: Delay associated with ACELP

Figure 4: The computational load distribution for MPMLQ

Figure 5: The computational load distribution for ACELP

Figure 6. Bit allocation for MPMLQ

Figure 7. Bit allocation for ACELP

[2] ITU-T Recommendation G.723.1, Dual Rate Speech Coder for Multimedia Communications Transmitting at 5.3 and 6.3 Kbps. March 1996.

[4] Andreas S. Spanias, “Speech Coding: A Tutorial Review,” Proc. IEEE, vol.82, no. 10, pp. 1541-1582, October 1994.

[5] P. Kroon and E. Deprettere. “A Class of Analysis-By-Synthesis Predictive Coders for High Quality Speech Coding at Rates Between 4.8 and 16 Kbps”, IEEE jour. On Selected Areas in Communications, pages 353-363, February 1988.

[6]K. Y. Lee. “Analysis By Synthesis Linear Predictive Coding”. PhD thesis, University of Surrey. Guildford, Surrey, U.K., October 1990.

[7] G.Frant “Acoustic theory of Speech Production.” Gravenhage, The Netherlands: Mounton and Co, 1960.

[8]Noman Junejo. “Study and Implementation of ITU-T G.723.1 in Java”. BE thesis, Hamdard Institute of Information Technology, Hamdard University Karachi, Pakistan., November 2002.

[9]G. avidson , ”Complexity reduction method for vector excitation coding,” in Proc. ICASSP-86 Tokyo, Japan, 1986, p.3055.

[11] R. W. Schafer and B. S. Atal. “Code Excited Linear Prediction High Quality Speech at Low Bit Rates”, Proc of ICASSP, pages 937-940, 1985.

[12] B. Atal, V. Cuperman, et al, “Advances in Speech Coding”, Norwell MA: Kluwer, 1990.

[14] A.Buzo et al.,”Speech coding based upon vector quantization,” IEEE Trans. Acoust., Speech, Signal Process., vol. ASSP-28, no 5, pp562-574, Oct. 1980.

[15]M.Akamine and K Miseki “CELP coding with an adaptive density pulse excitation model”, in Proc. ICASSP-90 New Mexico, April 1990, p.29

[16] S. Singhal and B. S. Atal. “Improving the performance of multi-pulse LPC coders at low bit rates”. In Proc. Of ICASSP, pages 1.3.1-1.3.4, San Diego, 1984.

[17] O. Fujimora, "Speech Coding and the Excitation Signal," IEEE Int. Comm. Conf. P

<<Home
		Noman Ahmed, A Q K Rajput, Naveed Ahmed Post Graduate Student Institute of Information Technology, Mehran University of Engineering and Technology *Professor Institute of Information Technology, Mehran University of Engineering and Technology
Heart patients click here

		Abstract Factors serving as constraints in today’s communication systems include bandwidth consumption, power, quality and delay. The desire to communicate effectively and exchange as much information as possible by transmitting only a few parameters, derives an entire industry. In Voice over Internet Protocol (VoIP) application, that require the transmission of speech, all these goals are addressed by development of efficient methods of speech compression. This paper compares two speech-encoding algorithms, differentiated by their fundamental approach to coding and decoding of speech signals. MPMLQ, Multi Pulse Maximum Likelihood Quantization, that separately encode the even and odd pulse positions by employing combinatorial coding. ACELP, Algebraic Code Excited Linear Prediction, is an enhanced version of CELP, Code Excited Linear Prediction speech coding algorithms, which employ an algebric codebook for efficient searching. These coding techniques will be compared based on the delay, computational load distribution, and no of bits transmitted for each technique, discussing the fundamentals behind each and concluding factors and result important to system designers, implementers and subscribers. Keywords MPMLQ, ACELP, ITU, Speech compression, CELP. Extended Summary Although with the emergence of optical fibers bandwidth in wired communications has become inexpensive still, there is a growing need for bandwidth conservation and enhanced privacy in wireless cellular and satellite communications. In particular, cellular communications have been enjoying a tremendous worldwide growth and there is a great deal of Research and Development (R&D) activity geared towards establishing global portable communications through wireless personal communication networks. On the other hand, there is a trend toward integrating voice-related applications (e.g., voicemail) on desktop and portable personal computers - often in the context of multimedia communications. Most of these applications require that the speech signal is in digital format so that it can be processed, stored, or transmitted under software control. This paper compares two of these speech coding techniques, based on the means by which they achieve compression: MPMLQ and ACELP. , Multi Pulse Maximum Likelihood Quantization, MPMLQ, and Algebric Code Excited Linear Prediction, ACELP, make an interesting pair because they addresses speech compression using two fundamentally different techniques, and both of these algorithms are part of the International Telecommunication Union (ITU) Recommendation G.723.1. Speech is generally band limited to 4 kHz (or 3.2 kHz) and sampled at 8kHz, although digital speech brings flexibility and opportunities for encryption, it is also associated (when uncompressed) with a high data rate and hence high requirements of transmission bandwidth and storage. Speech Coding or Speech Compression is the field concerned with obtaining compact digital representations of voice signals for the purpose of efficient transmission or storage. Speech coding involves sampling and amplitude quantization. The objective in speech coding is to represent speech with a minimum number of bits while maintaining its perceptual quality. Speech coders differ widely in their approaches to achieving signal compression. Based on the means by which they achieve compression: MPMLQ and ACELP are classified as analysis-by-synthesis coders. Analysis-by-synthesis speech coders are among the newest and most effective of modern speech coders The results demonstrate the delay, computational load distribution, and no of bits transmitted for each technique, obtained by testing our codec using the test vectors provided by ITU-T. The objective of this paper is to compare two fundamentally different speech-coding techniques, MPMLQ and ACELP implemented under ITU-T G.723.1 recommendation. We conclude form the results that the delay associated with MPMLQ is slightly greater than that of the ACELP. Next we analyzed the computation load distribution graphs, which shows the complexity level associated with each algorithm. It can be concluded form the graphs that MPMLQ stochastic codebook search takes 55% of the time where as ACELP stochastic codebook search takes 47% of the time. The bit allocation graphs demonstrate that ACELP takes less no of bits to transmit than MPMLQ. All these proven results make ACELP an attractive solution for low delay speech compression applications. However, the major tradeoff exists between the toll quality speech reconstructed using MPMLQ and the lower transmission rate and delay of ACELP technique, providing system designers, implementers and subscribers with flexibility. Figures Figure 1. Even and Odd pulse positions for MPMLQ Figure 2: Delay associated with MPMLQ Figure 3: Delay associated with ACELP Figure 4: The computational load distribution for MPMLQ Figure 5: The computational load distribution for ACELP Figure 6. Bit allocation for MPMLQ Figure 7. Bit allocation for ACELP References. [1] Gill Held, “Voice and Data Internetworking”, Osborne/McGraw-Hill, 2001. [2] ITU-T Recommendation G.723.1, Dual Rate Speech Coder for Multimedia Communications Transmitting at 5.3 and 6.3 Kbps. March 1996. [3] A.M Kondoz, Digital Speech, John Wiley & sons, January 2001. [4] Andreas S. Spanias, “Speech Coding: A Tutorial Review,” Proc. IEEE, vol.82, no. 10, pp. 1541-1582, October 1994. [5] P. Kroon and E. Deprettere. “A Class of Analysis-By-Synthesis Predictive Coders for High Quality Speech Coding at Rates Between 4.8 and 16 Kbps”, IEEE jour. On Selected Areas in Communications, pages 353-363, February 1988. [6]K. Y. Lee. “Analysis By Synthesis Linear Predictive Coding”. PhD thesis, University of Surrey. Guildford, Surrey, U.K., October 1990. [7] G.Frant “Acoustic theory of Speech Production.” Gravenhage, The Netherlands: Mounton and Co, 1960. [8]Noman Junejo. “Study and Implementation of ITU-T G.723.1 in Java”. BE thesis, Hamdard Institute of Information Technology, Hamdard University Karachi, Pakistan., November 2002. [9]G. avidson , ”Complexity reduction method for vector excitation coding,” in Proc. ICASSP-86 Tokyo, Japan, 1986, p.3055. [10] Luis Miguel Teixeira de Jesus, “Speech Coding and Synthesis Using Parametric Curvess”, S hesis, University of East Anglia., October 1997. [11] R. W. Schafer and B. S. Atal. “Code Excited Linear Prediction High Quality Speech at Low Bit Rates”, Proc of ICASSP, pages 937-940, 1985. [12] B. Atal, V. Cuperman, et al, “Advances in Speech Coding”, Norwell MA: Kluwer, 1990. [13]H.Abut, Ed., “Vector Quantization”. New York: IEEE Press 1990. [14] A.Buzo et al.,”Speech coding based upon vector quantization,” IEEE Trans. Acoust., Speech, Signal Process., vol. ASSP-28, no 5, pp562-574, Oct. 1980. [15]M.Akamine and K Miseki “CELP coding with an adaptive density pulse excitation model”, in Proc. ICASSP-90 New Mexico, April 1990, p.29 [16] S. Singhal and B. S. Atal. “Improving the performance of multi-pulse LPC coders at low bit rates”. In Proc. Of ICASSP, pages 1.3.1-1.3.4, San Diego, 1984. [17] O. Fujimora, "Speech Coding and the Excitation Signal," IEEE Int. Comm. Conf. P 49 ,1966. [18] David Flanagan, Java in a Nutshell, O’Reilly, August 1998. [19] Deitel & Deitel, Java how to program, Prentice Hall, 1999. [20] Cay. S & Gray Cornell, Core Java 2, SUN Microsystems,1999.






		<<Home


	Performance Comparison of MPMLQ and ACELP Speech Coding Algorithms