Experimenting with Lapped Transforms in Numerical Computation Libraries Using Polyphase Matrices and Strided Memory Views

Werner N, Edler B. Experimenting with Lapped Transforms in Numerical Computation Libraries Using Polyphase Matrices and Strided Memory Views. In: AES Convention 146. Audio Engineering Society; 2019. Paper 523. Available from: https://aes.org/publications/elibrary-page/?id=20381

@inproceedings{Werner2019_20381,
  author = {Werner, Nils and Edler, Bernd},
  title = {{Experimenting with Lapped Transforms in Numerical Computation Libraries Using Polyphase Matrices and Strided Memory Views}},
  booktitle = {AES Convention 146},
  note = {Paper 523},
  year = {2019},
  month = mar,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=20381}
}

TY  - CPAPER
TI  - Experimenting with Lapped Transforms in Numerical Computation Libraries Using Polyphase Matrices and Strided Memory Views
AU  - Werner, Nils
AU  - Edler, Bernd
T2  - AES Convention 146
M1  - Paper 523
PY  - 2019
DA  - 2019/03/06
UR  - https://aes.org/publications/elibrary-page/?id=20381
PB  - Audio Engineering Society
LA  - en
AB  - In this brief we present a framework for experimenting with lapped linear transforms in modern numerical computation libraries such as NumPy and Julia. We make use of the fact that these transforms can be represented as matrices (and oftentimes as sparse factorizations thereof), and that numerical computation libraries often support strided memory views. This strided memory view very elegantly solves the problem of processing several overlapping frames at once, while simultaneously allowing vectorization.
ER  -