Speech denoising using nonnegative matrix factorization with priors

@article{Wilson2008SpeechDU,
  title={Speech denoising using nonnegative matrix factorization with priors},
  author={Kevin W. Wilson and Bhiksha Raj and Paris Smaragdis and Ajay Divakaran},
  journal={2008 IEEE International Conference on Acoustics, Speech and Signal Processing},
  year={2008},
  pages={4029-4032}
}
  • K. Wilson, B. Raj, Ajay Divakaran
  • Published 12 May 2008
  • Computer Science, Physics
  • 2008 IEEE International Conference on Acoustics, Speech and Signal Processing
We present a technique for denoising speech using nonnegative matrix factorization (NMF) in combination with statistical speech and noise models. We compare our new technique to standard NMF and to a state-of-the-art Wiener filter implementation and show improvements in speech quality across a range of interfering noise types. 

Figures from this paper

Factorization With Temporal Dependencies for Speech Denoising
TLDR
This paper adds frame-to-frame regularization across time and shows that this additional regularization can also improve speech denoising results.
Recognize and separate approach for speech denoising using nonnegative matrix factorization
TLDR
This paper proposes a novel approach for denoising single-channel noisy speech signals using nonnegative matrix factorization (NMF), and indicates that recognizing the noise type from the mixed signal and using the corresponding specific noise dictionary provides better results than using a general noise dictionary in the NMF approach.
A new approach for speech enhancement based on a constrained Nonnegative Matrix Factorization
TLDR
The proposed scheme combines the noise Power Spectral Density estimation based on a constrained NMF and Wiener filtering to enhance the noisy speech to give smoother estimates of the nonnegative factors.
Speech Enhancement Using Sparse Convolutive Non-negative Matrix Factorization with Basis Adaptation
TLDR
It is demonstrated that an underlying speech representation which, in addition to applying sparsity, also adapts to the noisy acoustics improves overall enhancement quality.
Dual Channel Speech Denoising Based on Sparse Representation
TLDR
A novel speech denoising algorithm is proposed that firstly divides the noisy speech into a disjoint part and a joint part, and then removals the joint part which is mainly noise.
Learning speech features in the presence of noise: Sparse convolutive robust non-negative matrix factorization
  • R. FréinS. Rickard
  • Computer Science
    2009 16th International Conference on Digital Signal Processing
  • 2009
TLDR
It is shown that the proposed technique, namely Sparse convolutive robust non-negative matrix factorization, is robust in the presence of noise due to the explicit treatment of noise as an interfering source in the factorization.
Experiments on deep learning for speech denoising
TLDR
A very lightweight procedure that can predict clean speech spectra when presented with noisy speech inputs is proposed and it is concluded that it can perform better than some comparable single-channel approaches and that it is able to generalize well across various speakers, noise types and signal-to-noise ratios.
Speech enhancement using convolutive nonnegative matrix factorization with cosparsity regularization
TLDR
A novel method for speech enhancement based on Convolutive Non-negative Matrix Factorization (CNMF) improves sparse reconstruction of speech, especially in low SNR conditions.
Non-negative matrix factorization with linear constraints for single-channel speech enhancement
TLDR
It has been shown that the proposed NMF-based approach to the semi-supervised single-channel speech enhancement problem outperforms some of the state-of-the-art noise suppression techniques in terms of signal-to-noise ratio.
...
...

References

SHOWING 1-10 OF 11 REFERENCES
Sparse Overcomplete Decomposition for Single Channel Speaker Separation
TLDR
An algorithm for separating multiple speakers from a mixed single channel recording based on a model proposed by Raj and Smaragdis (2005) and a probabilistic framework to achieve sparsity is proposed.
Relation between PLSA and NMF and implications
TLDR
It is shown that PLSA solves the problem of NMF with KL divergence, and the implications of this relationship are explored.
From Learning Music to Learning to Separate
TLDR
It is shown that the same type of reasoning that allows one to find building elements of a musical audio stream can also be used to find and extract elements contained in auditory scenes.
Objective measures of speech quality
Algorithms for Non-negative Matrix Factorization
TLDR
Two different multiplicative algorithms for non-negative matrix factorization are analyzed and one algorithm can be shown to minimize the conventional least squares error while the other minimizes the generalized Kullback-Leibler divergence.
New Algorithms for Non-Negative Matrix Factorization in Applications to Blind Source Separation
  • A. CichockiR. ZdunekS. Amari
  • Computer Science
    2006 IEEE International Conference on Acoustics Speech and Signal Processing Proceedings
  • 2006
In this paper we develop several algorithms for non-negative matrix factorization (NMF) in applications to blind (or semi blind) source separation (BSS), when sources are generally statistically
Algorithms for nonnegative matrix factorization
  • NIPS
  • 2000
Darpa timit acoustic-phonetic continous speech corpus cd-rom documentation
  • Tech. Rep
  • 1993
Speech processing, transmission and quality aspects (stq); distributed speech recognition; advanced front-end feature extraction algorithm; compression algorithms
  • Speech processing, transmission and quality aspects (stq); distributed speech recognition; advanced front-end feature extraction algorithm; compression algorithms
  • 2003
Speech processing , transmission and quality aspects ( stq ) ; distributed speech recognition ; advanced front - end feature extraction algorithm ; compression algorithms
  • Tech . Rep . ETSI ES
  • 2003
...
...