首页> 外文期刊>IEEE Transactions on Information Theory >Two-stage vector quantization-lattice vector quantization
【24h】

Two-stage vector quantization-lattice vector quantization

机译:两阶段向量量化-点阵向量量化

获取原文
获取原文并翻译 | 示例
           

摘要

A two-stage vector quantizer is introduced that uses anunstructured first-stage codebook and a second-stage lattice codebook.Joint optimum two-stage encoding is accomplished by exhaustive search ofthe parent codebook of the two-stage product code. Due to the relativeease of lattice vector quantization, optimum encoding is feasible formoderate-to-large encoding rates and vector dimensions, provided thefirst-stage codebook size is kept reasonable. For memoryless Gaussianand Laplacian sources, encoding rates of 2 to 3 b/sample, and vectordimensions of 8 to 35 the signal-to-noise ratio performance iscomparable or superior to equivalent-delay encoding results previouslyreported. For Gaussian sources with memory, the effectiveness of theencoding method is dependent on the feasibility of using a large enoughfirst-stage vector quantizer codebook to exploit most of the sourcememory
机译:介绍了一种使用非结构化第一级码本和第二级格码本的两级矢量量化器。通过对两级乘积码的母码本进行穷举搜索,实现了最优的两级联合编码。由于晶格矢量量化的相对简便性,如果第一级码本的大小保持合理,则对于中等到较大的编码率和矢量尺寸,最佳编码是可行的。对于无记忆的高斯和拉普拉斯信号源,编码率为2到3 b /样本,矢量尺寸为8到35,信噪比性能可比或优于先前报告的等效延迟编码结果。对于具有内存的高斯源,编码方法的有效性取决于使用足够大的第一级矢量量化器代码簿来利用大部分源存储器的可行性。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

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