Nonlinear Residual Echo Suppression Based on Multi-stream Conv-TasNet

Chen, Hongsheng; Xiang, Teng; Chen, Kai; Lü, Jing

doi:10.48550/arxiv.2005.07631

Cited by 2 publications

(2 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These distortions can vary at different volumes, temperatures and between different loudspeakers. A common practice is to apply a functional non-linearity to mimic loudspeaker distortion as in [18]. Of course, the highest fidelity way to capture these effects in training data is to record echoed reference outputs in real rooms, but this has the considerable downside of being expensive and time consuming.…”

Section: Data Preparationmentioning

confidence: 99%

A Neural Acoustic Echo Canceller Optimized Using An Automatic Speech Recognizer And Large Scale Synthetic Data

Howard¹,

Park²,

Shabestary³

et al. 2021

Preprint

View full text Add to dashboard Cite

We consider the problem of recognizing speech utterances spoken to a device which is generating a known sound waveform; for example, recognizing queries issued to a digital assistant which is generating responses to previous user inputs. Previous work has proposed building acoustic echo cancellation (AEC) models for this task that optimize speech enhancement metrics using both neural network as well as signal processing approaches.Since our goal is to recognize the input speech, we consider enhancements which improve word error rates (WERs) when the predicted speech signal is passed to an automatic speech recognition (ASR) model. First, we augment the loss function with a term that produces outputs useful to a pre-trained ASR model and show that this augmented loss function improves WER metrics. Second, we demonstrate that augmenting our training dataset of real world examples with a large synthetic dataset improves performance. Crucially, applying SpecAugment style masks to the reference channel during training aids the model in adapting from synthetic to real domains. In experimental evaluations, we find the proposed approaches improve performance, on average, by 57% over a signal processing baseline and 45% over the neural AEC model without the proposed changes.

show abstract

Section: Data Preparationmentioning

confidence: 99%

A Neural Acoustic Echo Canceller Optimized Using An Automatic Speech Recognizer And Large Scale Synthetic Data

Howard¹,

Park²,

Shabestary³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Lately, the objective deep noise suppression mean opinion score (DNSMOS) metric has been proposed to estimate human ratings and has shown great accuracy [12]. Regarding the task of RES, speech quality during double-talk is traditionally evaluated using the objective signal-to-distortion ratio (SDR) metric [13], e.g., in [14][15][16][17][18][19]. Unfortunately, the SDR is affected by both desired-speech distortion and residual-echo presence, which renders it unreliable in predicting the DNSMOS and unreliable in predicting human perception of speech quality [12].…”

Section: Introductionmentioning

confidence: 99%

Objective Metrics to Evaluate Residual-Echo Suppression During Double-Talk

Ivry

Cohen

Berdugo

2021

2021 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA)

View full text Add to dashboard Cite

Human subjective evaluation is optimal to assess speech quality for human perception. The recently introduced deep noise suppression mean opinion score (DNSMOS) metric was shown to estimate human ratings with great accuracy. The signal-to-distortion ratio (SDR) metric is widely used to evaluate residual-echo suppression (RES) systems by estimating speech quality during doubletalk. However, since the SDR is affected by both speech distortion and residual-echo presence, it does not correlate well with human ratings according to the DNSMOS. To address that, we introduce two objective metrics to separately quantify the desiredspeech maintained level (DSML) and residual-echo suppression level (RESL) during double-talk. These metrics are evaluated using a deep learning-based RES-system with a tunable design parameter. Using 280 hours of real and simulated recordings, we show that the DSML and RESL correlate well with the DNSMOS with high generalization to various setups. Also, we empirically investigate the relation between tuning the RES-system design parameter and the DSML-RESL tradeoff it creates and offer a practical design scheme for dynamic system requirements.

show abstract

Nonlinear Residual Echo Suppression Based on Multi-stream Conv-TasNet

Cited by 2 publications

References 23 publications

A Neural Acoustic Echo Canceller Optimized Using An Automatic Speech Recognizer And Large Scale Synthetic Data

A Neural Acoustic Echo Canceller Optimized Using An Automatic Speech Recognizer And Large Scale Synthetic Data

Objective Metrics to Evaluate Residual-Echo Suppression During Double-Talk

Contact Info

Product

Resources

About