表1不同碼本生成算法譜失真比較
從表1數(shù)據(jù)可以看到,同是23bits的量化,聯(lián)合碼本設計MSVQ與應用SR算法生成碼本的MSVQ相比較,有大約1個比特的改善,接近于應用SR算法24bits量化的效果。甚至優(yōu)于文獻[2]中MSVQ算法的26bits量化(平均譜失真0.93dB)。平均譜失真為0.87dB,大于4dB的譜失真統(tǒng)計為0,達到了透明量化的要求。本文研究結(jié)果已經(jīng)成功應用于1/2kbps可變速率聲碼器項目中。
參考文獻
1 Itakura F. Line spectrum representation of linear predictive coefficients of speech signals [J]. J.Acoust.Soc.Amer.,1975; 57:S35
2 Paliwal K K, Atal B S. Efficient vector quantization of LPC parameters at 24 bits/frame [J]. Proc. ICASSP,1991:661～664
3 Juang B H, Gray A H. Multiple stage vector quantization for speech coding[J]. Proc.ICASSP,1982:597～600
4 Anderson J, Bodie J. Least square quantization in PCM [J]. IEEE Trans. Inform. Theory, 1975;IT-21:379～387
5 Chan W Y, Gupta S, Gersho A. Enhanced Multistage Vector Quantization by Joint Codebook Design [J]. IEEE Transations on Communications, 1992;40(11):1693～1697
6 Gardner W R, Rao B D. Theoretical Analysis of the High-Rate Vector Quantization of LPC Parameters [J]. IEEE Transactions on Speech and Audio Processing,1995;3(5):367～381
7 楊行峻, 遲惠生.語音信號數(shù)字處理 [M]. 北京:電子工業(yè)出版社, 1995
