Noise power spectral density estimation based on optimal smoothing and minimum statistics

By Rainer Martin. IEEE Transactions on Speech and Audio Processing, 2001

1,738 citations recorded by the source corpus.

Subfields: Computational Mechanics, Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/e59e303e8f8fca708fe557bc36babb254ffa07f8

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Speech enhancement for non-stationary noise environments (2001, background): “…density function SNR signal-to-noise ratio STFT short-time Fourier transform STSA short-time spectral amplitude However, this noise estimate is sensitive to outliers [24], generally biased [16], and its variance is about twice as large as the variance of a conventional noise estimator [15].”
A noise-estimation algorithm for highly non-stationary environments (2006, methodology): “Comparison between the noise spectrum (for f = 1 kHz) estimated using the proposed algorithm (thick line) and Martin (cid:1) s (Martin, 2001) (dashed line) algorithm for a sentence corrupted by car noise ( t < 1.8 s) followed by a sentence corrupted by multi-talker babble ( t > 1.8 s).”
Monaural speech segregation based on pitch tracking and amplitude modulation (2004, methodology): “For example, methods have been proposed to estimate the short-time spectra of interference [23] or to extract speech based Manuscript received August 2003; revised December 2003.”
Noise spectrum estimation in adverse environments: improved minima controlled recursive averaging (2003, methodology): “In contrast to the MS and related methods [5], [13], the smoothing of the noisy power spectrum is carried out in both time and frequency.”
DeepMMSE: A Deep Learning Approach to MMSE-Based Noise Power Spectral Density Estimation (2020, methodology): “In this subsection, the noise PSD tracking performance of the proposed DeepMMSE method is evaluated against five state-of-the-art noise PSD trackers, including the MS [14], MCRA-2 [17], MMSE [20], SPP [22], and ImMMSE [23] methods.”
Deep Learning for Environmentally Robust Speech Recognition (2017, methodology): “ethods is strongly affected by the noise modelling approach, more sophisticated noise estimation techniques have also been developed, for example, tracking spectral minima also during speech activity [53], [54]. 4 Whilst an instantaneous representation of the noise is obtained, noise reduction metho”
Unbiased MMSE-Based Noise Power Estimation With Low Complexity and Low Tracking Delay (2012, methodology): “In noise power estimation, it is a common assumption that the noise signal is more stationary than the speech signal [15].”
Speech enhancement based on minimum mean-square error estimation and supergaussian priors (2005, methodology): “To determine the variance of the slightly nonstationary car noise a minimum statistics [22] noise estimator is employed.”
An Algorithm for Intelligibility Prediction of Time–Frequency Weighted Noisy Speech (2011, methodology): “The noise PSD in EM and SG is estimated using minimum statistics [24] and the noise-tracker by Hendriks et al. [25], respectively.”
The TORGO database of acoustic and articulatory speech from speakers with dysarthria (2011, methodology): “Finally, acoustic noise reduction is performed using the minimum mean squared error estimate of the spectral amplitude (Ephraim and Malah 1985; Martin 2001).”
rVAD: An Unsupervised Segment-Based Robust Voice Activity Detection Method (2019, methodology): “7%, respectively. rVAD (MSNE-mod) performs marginally worse than the supervised BUT-VAD trained on the RATS database, but better than the supervised GMM-PLP-RASTA also trained on the RATS database.”
Supervised and Unsupervised Speech Enhancement Using Nonnegative Matrix Factorization (2013, background): “Furthermore, we included two state-of-the-art approaches in our experiments: The STSAGenGamma approach, identical to that of Subsection IV-A, and a Wiener filter in which the noise PSD was estimated using [12] and a decision-directed approach [50] was used to implement the filter.”
Robust signal-to-noise ratio estimation based on waveform amplitude distribution analysis (2008, methodology): “In another approach, Martin [4] used the low-energy envelope in frequency bands to estimate the SNR level.”