Auto-DSP: Learning to Optimize Acoustic Echo Cancellers

Casebeer, Jonah; Bryan, Nicholas J.; Smaragdis, Paris

doi:10.1109/waspaa52581.2021.9632678

Cited by 6 publications

(13 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…to reduce the dynamic range and facilitate training, but keep the phases unchanged. This pre-processing was found useful in several previous works [60], [69], although previous work used explicit clipping, which we found unnecessary.…”

Section: Optimizer Architecture and Lossmentioning

confidence: 93%

“…Such works, however, focus on creating learned optimizers for training neural networks in an offline setting, where the latter network is the final product, and the learned optimizer is otherwise discarded (or otherwise used to train additional networks). Moreover, this work has had little application to AFs, except for our own work [60], which we extend here.…”

Section: Introductionmentioning

confidence: 98%

“…In contrast, there are a small number of works that more tightly couple neural networks and AFs and use DNNs for optimal control of AFs. Recently, it was shown that DNNs can estimate statistics to control step-sizes [59] or estimate entire updates [60] for a single-channel AEC. Similarly, past work has used DNNs to predict updates for the internal statistics of multi-channel beamformers [61] and to learn source-models for multi-channel source separation [62].…”

Section: Introductionmentioning

confidence: 99%

“…3) We show how a single hyperparameter configuration of our approach, trained with different datasets and losses, can outperform all common AF baselines and/or several past state-of-the-art methods according to one or more evaluation metrics and are suitable for real-time operation on commodity hardware. Compared to our previous work on AEC in single-talk [60], we present several new core contributions including core algorithmic improvements and extensive validation across tasks. We release demos for each task and open source code 1 , including all baselines, creating an extensive resource for AFs.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Meta-AF: Meta-Learning for Adaptive Filters

Casebeer¹,

Bryan²,

Smaragdis³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Adaptive filtering algorithms are pervasive throughout modern society and have had a significant impact on a wide variety of domains including audio processing, telecommunications, biomedical sensing, astropyhysics and cosmology, seismology, and many more. Adaptive filters typically operate via specialized online, iterative optimization methods such as least-mean squares or recursive least squares and aim to process signals in unknown or nonstationary environments. Such algorithms, however, can be slow and laborious to develop, require domain expertise to create, and necessitate mathematical insight for improvement. In this work, we seek to go beyond the limits of human-derived adaptive filter algorithms and present a comprehensive framework for learning online, adaptive signal processing algorithms or update rules directly from data. To do so, we frame the development of adaptive filters as a metalearning problem in the context of deep learning and use a form of self-supervision to learn online iterative update rules for adaptive filters. To demonstrate our approach, we focus on audio applications and systematically develop meta-learned adaptive filters for five canonical audio problems including system identification, acoustic echo cancellation, blind equalization, multichannel dereverberation, and beamforming. For each application, we compare against common baselines and/or current state-ofthe-art methods and show we can learn high-performing adaptive filters that operate in real-time and, in most cases, significantly out perform all past specially developed methods for each task using a single general-purpose configuration of our method.

show abstract

Section: Optimizer Architecture and Lossmentioning

confidence: 93%

Section: Introductionmentioning

confidence: 98%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Meta-AF: Meta-Learning for Adaptive Filters

Casebeer¹,

Bryan²,

Smaragdis³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…This approach, however, requires expert knowledge, design trade-offs, and can be difficult or impossible to implement exactly. Existing audio effects implemented in this manner include infinite impulse response (IIR) filters [24], reverberation [25], echo cancellers [26], DJ transitions [27], and reverse engineered effects [28]. These approaches are related to the growing body of work focused on the construction of audio synthesis models with differentiable components, which now include additive [15], subtractive [29], waveshaping [30], and wavetable [31] synthesizers.…”

Section: Differentiable Audio Effectsmentioning

confidence: 99%

Style Transfer of Audio Effects with Differentiable Signal Processing

Steinmetz¹,

Bryan²,

Reiss³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

We present a framework that can impose the audio effects and production style from one recording to another by example with the goal of simplifying the audio production process. We train a deep neural network to analyze an input recording and a style reference recording, and predict the control parameters of audio effects used to render the output. In contrast to past work, we integrate audio effects as differentiable operators in our framework, perform backpropagation through audio effects, and optimize end-to-end using an audio-domain loss. We use a self-supervised training strategy enabling automatic control of audio effects without the use of any labeled or paired training data. We survey a range of existing and new approaches for differentiable signal processing, showing how each can be integrated into our framework while discussing their trade-offs. We evaluate our approach on both speech and music tasks, demonstrating that our approach generalizes both to unseen recordings and even to sample rates different than those seen during training. Our approach produces convincing production style transfer results with the ability to transform input recordings to produced recordings, yielding audio effect control parameters that enable interpretability and user interaction.

show abstract