Scholarlink

Block transforms (e.g., the DFT) can be viewed as critically sampled uniform filter banks. Each transform coefficient is a downsampled subband sample, with the analysis/synthesis “filters” given directly by columns/rows of the transform and its inverse. This chapter shows how to read off the equivalent impulse responses, explains why perfect reconstruction (PR) follows from transform invertibility, and illustrates the interpretation with Python examples. We also comment on windowing and overlap (WOLA/MDCT) as practical ways to improve per-channel selectivity while retaining PR.

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

Block Transforms as Filter Banks

  • 章节

  • 发表日期：

    2026年

查看全文
  • Gerald Schuller

摘要

Block transforms (e.g., the DFT) can be viewed as critically sampled uniform filter banks. Each transform coefficient is a downsampled subband sample, with the analysis/synthesis “filters” given directly by columns/rows of the transform and its inverse. This chapter shows how to read off the equivalent impulse responses, explains why perfect reconstruction (PR) follows from transform invertibility, and illustrates the interpretation with Python examples. We also comment on windowing and overlap (WOLA/MDCT) as practical ways to improve per-channel selectivity while retaining PR.