Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Scaling Up Adaptive Filter Optimizers

Jonah Casebeer, Nicholas J. Bryan, Paris Smaragdis

TL;DR

This work introduces a new online adaptive filtering method called supervised multi-step adaptive filters (SMS-AF), which uses neural networks to control or optimize linear multi-delay or multi-channel frequency-domain filters and can flexibly scale-up performance at the cost of increased compute.

Abstract

We introduce a new online adaptive filtering method called supervised multi-step adaptive filters (SMS-AF). Our method uses neural networks to control or optimize linear multi-delay or multi-channel frequency-domain filters and can flexibly scale-up performance at the cost of increased compute -- a property rarely addressed in the AF literature, but critical for many applications. To do so, we extend recent work with a set of improvements including feature pruning, a supervised loss, and multiple optimization steps per time-frame. These improvements work in a cohesive manner to unlock scaling. Furthermore, we show how our method relates to Kalman filtering and meta-adaptive filtering, making it seamlessly applicable to a diverse set of AF tasks. We evaluate our method on acoustic echo cancellation (AEC) and multi-channel speech enhancement tasks and compare against several baselines on standard synthetic and real-world datasets. Results show our method performance scales with inference cost and model capacity, yields multi-dB performance gains for both tasks, and is real-time capable on a single CPU core.

Scaling Up Adaptive Filter Optimizers

TL;DR

This work introduces a new online adaptive filtering method called supervised multi-step adaptive filters (SMS-AF), which uses neural networks to control or optimize linear multi-delay or multi-channel frequency-domain filters and can flexibly scale-up performance at the cost of increased compute.

Abstract

We introduce a new online adaptive filtering method called supervised multi-step adaptive filters (SMS-AF). Our method uses neural networks to control or optimize linear multi-delay or multi-channel frequency-domain filters and can flexibly scale-up performance at the cost of increased compute -- a property rarely addressed in the AF literature, but critical for many applications. To do so, we extend recent work with a set of improvements including feature pruning, a supervised loss, and multiple optimization steps per time-frame. These improvements work in a cohesive manner to unlock scaling. Furthermore, we show how our method relates to Kalman filtering and meta-adaptive filtering, making it seamlessly applicable to a diverse set of AF tasks. We evaluate our method on acoustic echo cancellation (AEC) and multi-channel speech enhancement tasks and compare against several baselines on standard synthetic and real-world datasets. Results show our method performance scales with inference cost and model capacity, yields multi-dB performance gains for both tasks, and is real-time capable on a single CPU core.
Paper Structure (21 sections, 7 equations, 1 figure, 2 tables, 1 algorithm)

This paper contains 21 sections, 7 equations, 1 figure, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. Adaptive Filters
  4. Adaptive Filter Optimizers
  5. Learned Optimizers
  6. Scaling Up Learned Optimizers
  7. Scaling Up Feature Quality: Feature Pruning
  8. Scaling Up Supervision: Supervised Loss
  9. Scaling Up Feedback: Multi-Step Optimization
  10. Modifications to Overlap-Save for AEC
  11. Perspectives
  12. Experimental Design
  13. Experiments
  14. AEC Configuration
  15. GSC Configuration
  16. ...and 6 more sections

Figures (1)

  • Figure 1: Acoustic echo cancellation performance vs. model size, optimization steps per time-frame (opt. steps), and supervision levels. Bubble size shows real-time-factor (RTF) where smaller is faster, inner-shape shows opt. steps, and the vertical dotted line separates unsupervised (left) and supervised (right) approaches. SMS-AF, in bold on the far right, demonstrate robust scaling performance in terms of parameters, and RTF.