Acoustic Echo Cancellation Using Deep Complex Neural Network with Nonlinear Magnitude Compression and Phase Information

Peng, Renhua; Cheng, Linjuan; Zheng, Chengshi; Li, Xiaodong

doi:10.21437/interspeech.2021-2022

Cited by 22 publications

(9 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As the delay between the far-end (FE) and near-end (NE) signal is generally unknown, in this work we propose employing dynamic delay compensation (DDC) via applying the GCC-PHAT algorithm [18] to find the time delay between the two signals and to then delay the far-end signal by the estimated time delay. Previous studies [5,8,9] have reported that a time alignment pre-processing stage helps to improve the AEC performance and to increase the robustness of the subsequent processing stages against delayed echo signals.…”

Section: Dynamic Delay Compensation (Ddc)mentioning

confidence: 99%

“…where J xC ℓ represents either (8) or (9). We then combine both J mC and J cC to the final loss (with β = 0.7, chosen based on the optimal performance for the trained models achieved on D test,MS syn )…”

Section: Training: Wideband Aec and Pfmentioning

confidence: 99%

“…Deep AEC studies can be divided into two groups: singlestage methods [3][4][5] naturally prohibiting a separate loss for AEC and residual echo suppression PF, and multi-stage methods [6][7][8][9][10], which consist of an AEC stage to remove major parts of the echo followed by a DNN PF to suppress residual echo and noise. Multistage methods can be further divided into two subgroups: Either (i) both AEC and PF are DNNs (DNN AEC/PF) [10], or (ii) a linear AEC [11,12] is followed by a DNN PF (LAEC/DNN PF) [6][7][8][9].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Bandwidth-Scalable Fully Mask-Based Deep FCRN Acoustic Echo Cancellation and Postfiltering

Seidel¹,

Olsson²,

Haddad³

et al. 2022

Preprint

View full text Add to dashboard Cite

Although today's speech communication systems support various bandwidths from narrowband to super-wideband and beyond, stateof-the art DNN methods for acoustic echo cancellation (AEC) are lacking modularity and bandwidth scalability. Our proposed DNN model builds upon a fully convolutional recurrent network (FCRN) and introduces scalability over various bandwidths up to a fullband (FB) system (48 kHz sampling rate). This modular approach allows joint wideband (WB) pre-training of mask-based AEC and postfilter stages with dedicated losses, followed by separate trainings of them on FB data. A third lightweight blind bandwidth extension stage is separately trained on FB data, flexibly allowing to extend the WB postfilter output towards higher bandwidths until reaching FB. Thereby, higher frequency noise and echo are reliably suppressed. On the ICASSP 2022 Acoustic Echo Cancellation Challenge blind test set we report a competitive performance, showing robustness even under highly delayed echo and dynamic echo path changes.

show abstract

Section: Dynamic Delay Compensation (Ddc)mentioning

confidence: 99%

Section: Training: Wideband Aec and Pfmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Bandwidth-Scalable Fully Mask-Based Deep FCRN Acoustic Echo Cancellation and Postfiltering

Seidel¹,

Olsson²,

Haddad³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Valin et al [6] combine a multi-delay block frequency-domain (MDF) filter and a PercepNet model for joint residual echo suppression and noise reduction. Peng et al [7] introduce a time delay estimation (TDE) block besides the adaptive filter to alleviate the task difficulty and propose a gated complex convolutional recurrent neural network (GCCRN) as a post-filter. Zhang et al [8] apply a doubletalk friendly weighted recursive least square (wRLS) filter [9] and use a multi-task gated convolutional frequency-time-LSTM neural network (GFTNN) that predicts the near-end speech activity at the same time.…”

Section: Introductionmentioning

confidence: 99%

“…In this paper, we first propose a backbone gated temporal convolutional neural network termed as GTCNN for unconditional AEC. GTCNN is developed on the basis of GC-CRN [7], a competitive entry in the INTERSPEECH 2021 AEC-Challenge. By introducing a gated temporal convolution mechanism, GTCNN achieves state-of-the-art performance with fewer parameters.…”

Section: Introductionmentioning

confidence: 99%

Personalized Acoustic Echo Cancellation for Full-duplex Communications

Zhang¹,

Wang²,

Yukai³

et al. 2022

Preprint

View full text Add to dashboard Cite

Deep neural networks (DNNs) have shown promising results for acoustic echo cancellation (AEC). But the DNN-based AEC models let through all near-end speakers including the interfering speech. In light of recent studies on personalized speech enhancement, we investigate the feasibility of personalized acoustic echo cancellation (PAEC) in this paper for full-duplex communications, where background noise and interfering speakers may coexist with acoustic echoes. Specifically, we first propose a novel backbone neural network termed as gated temporal convolutional neural network (GTCNN) that outperforms stateof-the-art AEC models in performance. Speaker embeddings like d-vectors are further adopted as auxiliary information to guide the GTCNN to focus on the target speaker. A special case in PAEC is that speech snippets of both parties on the call are enrolled. Experimental results show that auxiliary information from either the near-end speaker or the far-end speaker can improve the DNN-based AEC performance. Nevertheless, there is still much room for improvement in the utilization of the finitedimensional speaker embeddings.

show abstract

Neuralecho: Hybrid of Full-Band and Sub-Band Recurrent Neural Network For Acoustic Echo Cancellation and Speech Enhancement

Yu,

Xu,

Zhang

et al. 2023

2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)

View full text Add to dashboard Cite

Acoustic Echo Cancellation Using Deep Complex Neural Network with Nonlinear Magnitude Compression and Phase Information

Cited by 22 publications

References 7 publications

Bandwidth-Scalable Fully Mask-Based Deep FCRN Acoustic Echo Cancellation and Postfiltering

Bandwidth-Scalable Fully Mask-Based Deep FCRN Acoustic Echo Cancellation and Postfiltering

Personalized Acoustic Echo Cancellation for Full-duplex Communications

Neuralecho: Hybrid of Full-Band and Sub-Band Recurrent Neural Network For Acoustic Echo Cancellation and Speech Enhancement

Contact Info

Product

Resources

About