Abstract
We introduce a general and data-driven adaptation-control framework for stereophonic acoustic-echo cancellation. The adaptation update rule for the filters that estimate the actual echo paths is compactly expressed with the widely-linear model in the complex time domain. A single step-size parameter that governs the behavior of the adaptation process is optimized by minimizing the misalignment between the actual echo paths and their filtered estimate. The relation between acoustic signals and the optimal step-size is learned via a deep neural network. In test mode, the optimal step-size prediction is inferred by the network and fed to the sign-error nor-malized least mean-squares (SNLMS) adaptive filter for echo-paths tracking. Real and simulated data show advantageous performance in single and double-talk scenarios across various acoustic setups.
Original language | English |
---|---|
Title of host publication | Proceedings of the 2023 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2023 |
Publisher | Institute of Electrical and Electronics Engineers Inc. |
ISBN (Electronic) | 9798350323726 |
DOIs | |
State | Published - 2023 |
Externally published | Yes |
Event | 2023 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2023 - New Paltz, United States Duration: 22 Oct 2023 → 25 Oct 2023 |
Publication series
Name | IEEE Workshop on Applications of Signal Processing to Audio and Acoustics |
---|---|
Volume | 2023-October |
ISSN (Print) | 1931-1168 |
ISSN (Electronic) | 1947-1629 |
Conference
Conference | 2023 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA 2023 |
---|---|
Country/Territory | United States |
City | New Paltz |
Period | 22/10/23 → 25/10/23 |
Bibliographical note
Publisher Copyright:© 2023 IEEE.
Keywords
- Stereophonic acoustic echo cancellation
- adaptation control
- deep learning
- sign-error NLMS
- variable step-size