This chapter extends the transform-as-filter-bank view to the discrete cosine transform (DCT), with emphasis on the DCT-IV used widely in audio coding. We extract the equivalent analysis/synthesis impulse responses from the DCT-IV matrix and discuss their limited frequency selectivity (rectangular time support). To overcome these limits while retaining critical sampling and perfect reconstruction (PR), we introduce the polyphase representation. By reorganizing filtering and sampling into vector–polynomial products, polyphase turns “filter+decimate” into a low-rate matrix product in the z-domain and yields concise PR conditions.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

DCT and Polyphase Representation

  • Gerald Schuller

摘要

This chapter extends the transform-as-filter-bank view to the discrete cosine transform (DCT), with emphasis on the DCT-IV used widely in audio coding. We extract the equivalent analysis/synthesis impulse responses from the DCT-IV matrix and discuss their limited frequency selectivity (rectangular time support). To overcome these limits while retaining critical sampling and perfect reconstruction (PR), we introduce the polyphase representation. By reorganizing filtering and sampling into vector–polynomial products, polyphase turns “filter+decimate” into a low-rate matrix product in the z-domain and yields concise PR conditions.