Constrained Convolutional-Recurrent Networks to Improve Speech Quality with Low Impact on Recognition Accuracy

Fakoor, Rasool; He, Xiaodong; Tashev, Ivan; Zarar, Shuayb

doi:10.1109/icassp.2018.8462042

Cited by 7 publications

(4 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multi-task learning for audio quality. Multi-task learning (MTL) [30] has been beneficial to many speech applications [31][32][33][34]. MTL aims to leverage useful information contained in multiple related tasks to help improve the generalization performance on all tasks.…”

Section: Related Workmentioning

confidence: 99%

NORESQA: A Framework for Speech Quality Assessment using Non-Matching References

Manocha,

Xu,

Kumar

2021

Preprint

View full text Add to dashboard Cite

The perceptual task of speech quality assessment (SQA) is a challenging task for machines to do. Objective SQA methods that rely on the availability of the corresponding clean reference have been the primary go-to approaches for SQA. Clearly, these methods fail in real-world scenarios where the ground truth clean references are not available. In recent years, non-intrusive methods that train neural networks to predict ratings or scores have attracted much attention, but they suffer from several shortcomings such as lack of robustness, reliance on labeled data for training and so on. In this work, we propose a new direction for speech quality assessment. Inspired by human's innate ability to compare and assess the quality of speech signals even when they have non-matching contents, we propose a novel framework that predicts a subjective relative quality score for the given speech signal with respect to any provided reference without using any subjective data. We show that neural networks trained using our framework produce scores that correlate well with subjective mean opinion scores (MOS) and are also competitive to methods such as DNSMOS [1], which explicitly relies on MOS from humans for training networks. Moreover, our method also provides a natural way to embed quality-related information in neural networks, which we show is helpful for downstream tasks such as speech enhancement.Preprint. Under review.

show abstract

Section: Related Workmentioning

confidence: 99%

NORESQA: A Framework for Speech Quality Assessment using Non-Matching References

Manocha,

Xu,

Kumar

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Operating in noisy environments affect the performance of speech processing systems that are normally designed based on clean speech signals [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16]. Speech denoising involves the reduction or removal of the noisy part of the speech+noise mixture.…”

Section: Introductionmentioning

confidence: 99%

A review of multi-objective deep learning speech denoising methods

Azarang

Kehtarnavaz

2020

Speech Communication

View full text Add to dashboard Cite

“…More recently, a type of advanced network named convolutional recurrent neural networks (CRNN) is proposed, which takes advantages of both CNNs and RNNs. It is shown that it can obtain a lower word error rate (WER) in speech recognition tasks [20] and later it is introduced into speech enhancement tasks [21,22]. Compared with the single-type network, experiments have shown that the CRNN obtains better speech enhancement results.…”

Section: Introductionmentioning

confidence: 99%

Convolutional Recurrent Neural Network Based Progressive Learning for Monaural Speech Enhancement

Li,

Yuan,

Zheng

et al. 2019

Preprint

View full text Add to dashboard Cite

Recently, progressive learning has shown its capacity of improving speech quality and speech intelligibility when it is combined with deep neural network (DNN) and long short-term memory (LSTM) based monaural speech enhancement algorithms, especially in low signal-to-noise ratio (SNR) conditions. Nevertheless, due to a large number of parameters and highly computational complexity, it is hard to implement in current resource-limited micro-controllers and thus, it is important to significantly reduce both the amount of parameters and the computational load for practical applications. For this purpose, we propose a novel progressive learning framework with convolutional recurrent neural networks called PL-CRNN, which takes advantages of both convolutional neural networks and recurrent neural networks to drastically reduce the amount of parameters and simultaneously improve speech quality and speech intelligibility.Numerous experiments verify the effectiveness of proposed PL-CRNN model and indicate that it yields consistent better performance than the PL-DNN and PL-LSTM algorithms and also it gets results close even better than the CRNN in terms of various evaluation metrics. Compared with PL-DNN, PL-LSTM and state-of-the-art CRNN models, the proposed PL-CRNN algorithm can reduce the amount of parameters up to 77%, 93% and 93%, respectively.

show abstract

Constrained Convolutional-Recurrent Networks to Improve Speech Quality with Low Impact on Recognition Accuracy

Cited by 7 publications

References 13 publications

NORESQA: A Framework for Speech Quality Assessment using Non-Matching References

NORESQA: A Framework for Speech Quality Assessment using Non-Matching References

A review of multi-objective deep learning speech denoising methods

Convolutional Recurrent Neural Network Based Progressive Learning for Monaural Speech Enhancement

Contact Info

Product

Resources

About