Deep Adaptive Aec: Hybrid of Deep Learning and Adaptive Acoustic Echo Cancellation

Zhang, Hao; Kandadai, Srivatsan; Rao, H.I.K.; Kim, Minje; Pruthi, Tarun; Kristjansson, Trausti T.

doi:10.1109/icassp43922.2022.9746039

Cited by 22 publications

(15 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The PESQ is improved by 0.05 and WER is relatively improved by 11.8%. The results presented in [8] show that NLM-SNet substantially outperforms the fully DNN-based method, which is different from what we obtained here. The major reason is that the signals used for training and testing in [8] are recorded in scenarios with continuous echo path changing, which helps show the benefits of using hybrid methods.…”

Section: Nonlinear Transition Function T(•)contrasting

confidence: 99%

“…Such methods usually treat AEC as a source separation problem and directly estimate the nearend signal based on the microphone and far-end reference signal. While achieving good performance in general, DNN-based methods have shown limited utility in dealing with continuously changing echo paths [8]. In recent AEC challenges [9], two-stage hybrid systems [10,11,12] that use DNN as a nonlinear post-processor of a DSP-based adaptive filtering algorithm have shown promising results.…”

Section: Introductionmentioning

confidence: 99%

“…In such hybrid systems, DNNs perform nonlinear residual echo suppression, which compensates for the drawbacks of adaptive filtering algorithms. To further leverage the advantages of DNN and adaptive filtering algorithms, methods such as Deep Adaptive AEC [8] train DNNs where a linear adaptive algorithm is embedded as differentiable layers. Such a hybrid system has shown superior results in modeling a continuously changing echo path compared to DNN-based and two-stage hybrid methods.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

NeuralKalman: A Learnable Kalman Filter for Acoustic Echo Cancellation

Zhang¹,

Meng²,

Zhang³

et al. 2023

Preprint

View full text Add to dashboard Cite

The Kalman filter is widely used for addressing acoustic echo cancellation (AEC) problems due to their robustness to double-talk and fast convergence. However, the inability to model nonlinearity and the need to tune control parameters cast limitations on such adaptive filtering algorithms. In this paper, we integrate the frequency domain Kalman filter (FDKF) and deep neural networks (DNNs) into a hybrid method, called NeuralKalman, to leverage the advantages of deep learning and adaptive filtering algorithms. Specifically, we employ a DNN to estimate nonlinearly distorted far-end signals, a transition factor, and the nonlinear transition function in the state equation of the FDKF algorithm. Experimental results show that the proposed NeuralKalman improves the performance of FDKF significantly and outperforms strong baseline methods.

show abstract

Section: Nonlinear Transition Function T(•)contrasting

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

NeuralKalman: A Learnable Kalman Filter for Acoustic Echo Cancellation

Zhang¹,

Meng²,

Zhang³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Moreover, we deploy learned optimizers to solve AF tasks as the endgoal and do not use them to train downstream neural networks. We also note recent work that uses a supervised DNN to control the step-size of a D-KF for AEC [61] and another that uses a supervised DNN to predict both the step-size and a nonlinear reference signal for AEC [62]. Compared to these, we replace the entire update with a neural network, do not need supervisory signals, and investigate many tasks.…”

Section: E Related Workmentioning

confidence: 99%

“…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 [61], [62] or estimate entire updates [63] for a single-channel AEC. Similarly, past work has used DNNs to predict updates for the internal statistics of multi-channel beamformers [64] and to learn source-models for multi-channel source separation [65].…”

Section: Introductionmentioning

confidence: 99%

Meta-AF: Meta-Learning for Adaptive Filters

Casebeer

Bryan

Smaragdis

2023

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

Adaptive filtering algorithms are pervasive throughout signal processing and have had a material impact on a wide variety of domains including audio processing, telecommunications, biomedical sensing, astrophysics 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 improve upon hand-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, multi-channel dereverberation, and beamforming. We compare our approach against common baselines and/or recent state-ofthe-art methods. We show we can learn high-performing adaptive filters that operate in real-time and, in most cases, significantly outperform each method we compare against -all using a single general-purpose configuration of our approach.

show abstract