首页> 外国专利> Complexity Reduction Algorithm of QCELP Speech Coder

Complexity Reduction Algorithm of QCELP Speech Coder

机译:QCELP语音编码器的复杂度降低算法

摘要

The present invention is a method of cyclically using a previously calculated value by avoiding a complicated formula having a large amount of calculation used to calculate the energy E yy L and E yyl obtained in the search of pitch and codebook parameters in the QECLP algorithm. The goal of this method is to implement a cheaper and less MIPS chip for QCELP vocoder implementation. The configuration of the present invention reduces the complexity of calculating E yyL , which is an energy item found in the pitch parameter search in the QCELP vocoder algorithm.;E yyL = y L (n) y L (n) + S L (n), where S L (n) = S Ll (n) + y L2 (N h ) -y Ll2 (L p -l);Where L is pitch delay, y L (n) is the synthesized speech signal with pitch delay L, N h is the number of samples used to approximate the impulse response of the weighted synthesis filter, and L p is the number of samples in the pitch subframe ).;In addition, the computational complexity is reduced to calculate E yyl , which is an energy item that is found when searching codebook parameters in the QCELP vocoder algorithm.;E yyl = y L (n) y L (n) + S L (n), where S l (n) = S ll (n) + y l2 (N h ) -y ll2 (L c -l);Where I is the codebook index y l (n) is the synthesized speech signal with codebook index I, N h is the number of samples used to approximate the impulse response of the weighted synthesis filter, and L c is the sample of the todbook subframe Number).;In addition, when the selected codebook value becomes zero during the codebook parameter search, the complexity of calculating the energy item E yyl is reduced E yyl = E yyl-l -y l2 (L c -1), where I is the codebook index. y l (n) is the synthesized speech signal having codebook index I, N h is the number of samples used to approximate the impulse response of the weighted synthesis filter, and L c is the number of samples of the todbook subframe.;The effect of the present invention is to reduce the amount of computation by improving the high-energy energy equation calculated during the search of pitch and codebook parameters in the QCELP vocoder algorithm, and to implement the QCELP vocoder using the reduced complexity equation, it is cheaper and more MIPS. The effect is to make a vocoder chip with less.
机译:本发明是一种通过避免使用复杂的公式循环使用先前计算的值的方法,该复杂的公式具有用于计算能量E yy L 和E yyl 。此方法的目标是为QCELP声码器实现实现更便宜,更便宜的MIPS芯片。本发明的配置降低了计算E yyL 的复杂性,E yyL 是在QCELP声码器算法中的音调参数搜索中找到的能量项。 L (n)y L (n)+ S L (n),其中S L (n) = S Ll (n)+ y L 2 (N h )-y Ll 2 (L p -l);其中L是音调延迟,y L (n)是具有音调延迟的合成语音信号L,N h 是用于近似加权合成滤波器的脉冲响应的样本数,L p 是基音子帧中的样本数。另外,降低了计算复杂度以计算E yyl ,这是在QCELP声码器算法中搜索码本参数时发现的能量项。E yyl = y L (n)y L (n)+ S L (n),其中S l (n) = S <子> l l (n)+ y l 2 (N h )-y ll 2 (L c -l);其中I是密码本索引y l (n)是密码本索引为I的合成语音信号,N h 是用于近似加权合成滤波器的脉冲响应的样本数,而L c 是todbook子帧编号的样本)。在密码本参数搜索期间,密码本值变为零,从而降低了计算能量项E yyl 的复杂度E yyl = E yyl-1 - y l 2 (L c -1),其中I是密码本索引。 y l (n)是码本索引为I的合成语音信号,N h 是用于近似加权合成滤波器的脉冲响应的样本数,L c 是todbook子帧的样本数。本发明的效果是通过改进在搜索音调和码本参数期间计算出的高能能量方程来减少计算量。 QCELP声码器算法,并使用降低的复杂度方程实现QCELP声码器,它更便宜,而且具有更多的MIPS。效果是使声码器芯片更少。

著录项

  • 公开/公告号KR960024914A

    专利类型

  • 公开/公告日1996-07-20

    原文格式PDF

  • 申请/专利权人 양승택;

    申请/专利号KR19940036574

  • 发明设计人 장석진;이인성;김재원;

    申请日1994-12-24

  • 分类号G06F9/00;

  • 国家 KR

  • 入库时间 2022-08-22 03:44:50

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号