A spectral subtraction method for the enhancement of speech corrupted by nonwhite, nonstationary noise

McOlash, S.; Niederjohn, Russell J.

doi:10.1109/iecon.1995.483843

Cited by 13 publications

(7 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Consider a mixed speech signal composed of two parts ( ) which are independently generated by two sound sources. There is (4) Above is an energy bounded noise vector obeying a multi-dimensional Gaussian distribution. Generally and are sparse with respect to two different basis matrixes and ( and are integers larger than ), respectively.…”

Section: B the Relaxing Strategy For Articulation-oriented Restorationmentioning

confidence: 99%

“…In early implementations, a spectral subtraction approach was widely used to suppress Gaussian noise. This approach estimates the power spectral density (PSD) of a clean signal by subtracting the PSD of Gaussian noise from the PSD of the noisy signal [3], [4]. The estimation of PSD is performed within short time segments, because the short-time spectral amplitude carries important information about both speech quality and intelligibility.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Robust Time-frequency Decomposition Model for Suppression of Mixed Gaussian-impulse Noise in Audio Signals

Tong

Zhou

Zhang

et al. 2014

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

In this paper, we propose a robust time-frequency decomposition (RTFD) model to restore audio signals degraded by sparse impulse noise mixed with small dense Gaussian noise. This kind of noise is very common especially in old-time recordings. The proposed RTFD model is based on the observation that these degraded audio signals mainly contain four parts, i.e., the quasi-periodic and voiced part, the aperiodic and transient part, the arbitrarily large impulse noise and the small dense Gaussian noise. Sparsity and local correlations of corresponding parts are exploited to solve the RTFD model. We also heuristically develop a discriminative orthogonal matching pursuit (DOMP) algorithm to more precisely estimate sparse representing vectors. Specifically, the DOMP algorithm divides the whole atom set into two subsets, i.e., the active subset and the passive subset. Atoms in two subsets are treated discriminatively since sparsity regularization terms are not equally weighted. Based on RTFD and DOMP, we have developed two algorithms, i.e., the fidelity-oriented algorithm and the articulation-oriented algorithm. The proposed algorithms achieve considerable performance on both synthetic and real noisy signals. Results show that the articulation-oriented algorithm using DOMP obviously outperforms other algorithms in heavier impulse noise situations.Index Terms-Articulation-oriented, degraded, discriminative orthogonal matching pursuit (DOMP), Gaussian-impulse noise, restoration, robust time-frequency decomposition (RTFD).

show abstract

Section: B the Relaxing Strategy For Articulation-oriented Restorationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Robust Time-frequency Decomposition Model for Suppression of Mixed Gaussian-impulse Noise in Audio Signals

Tong

Zhou

Zhang

et al. 2014

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…The estimated background noise spectrum is then subtracted from the spectrum of the signal. This approach is analogous to speech background noise subtraction methods that rely on nonspeech intervals [14][15]. Thus, a general expression is defined using a flexible constant γ to get a subtracted spectrum,…”

Section: Displacementmentioning

confidence: 99%

Quantifying Human Indoor Activity Using a Software Radio-Based Radar

Godana

Leus

Barroso

2010

2010 First International Conference on Sensor Device Technologies and Applications

View full text Add to dashboard Cite

Human activity quantification consists of computing a numerical or qualitative metric that indicates the amount of movement a person engaged in a given time interval. Such a metric has important applications in elderly care, wellness and healthcare given the strong empirical relation between a person's health and his or her activity level. This paper proposes and evaluates methods to quantify the level of human activity in an indoor environment using a continuous wave radar. An experimental evaluation is carried out using a flexible and low-cost software defined radar platform. Results showed a good correlation between the proposed metrics and the motion sequence performed by the subject suggesting that accurate activity quantification in indoor environments can be achieved using a few simple off-body sensors.

show abstract

“…denotes the expectation operator[31]. Since the noise d C n) is assumed to have zero mean and is assumed to be uncorrelated withsCn) , the terms E[ SwCej{J)).Dw*Cej{J)) ] and E[ Sw*Cej{J)).DwCej{J)) ] reduce to zero.Therefore equation (5.13) can be written as[29],[31]:From above, the estimated short-time power spectrum of the clean signal is then given as[29]:or 1 Sw(e iOJ ) 12 = 1 Yw(e iOJ ) 12 -E[I Dw(e iOJ ) 12]1 Sw{e jOJ ) 12 = 1 Yw{e jOJ ) 12 -I Dw{e jOJ )12…”

mentioning

confidence: 99%

Enhancement of CN Tower lightning current derivative signals using a modified power spectral subtraction method

Mehmud¹

2021

Preprint

View full text Add to dashboard Cite

Lightning current measurements are possible using instrumental tall structures or rocket-triggered lightning. The CN Tower has been a source of lightning current data for the past 15 years. A major portion of research on the natural lightning is focused on developing lightning protection systems, and in order to do so, an accurate knowledge of the characteristics of lightning, including the return-stroke current, is required. The CN Tower is a transmission tower and it is expected that the recorded lightning current signals be corrupted with different kinds of noise. This makes it difficult to extract the return-stroke current waveform parameters (peak, 10-90% rise-time to peak, maximum steepness, pulse width etc.) from the measured waveforms. In this project, an over-subtraction and residual noise reduction based power spectral subtraction method has been developed in order to de-noise the lighting return-stroke current derivative signals measured at the CN Tower. In order to evaluate the proposed de-noising technique, the derivative of Heidler function is used to model the measured return-stroke current derivative signal. The measured current derivative signal is simulated using the Heidler derivative model after artificially corrupting it with noise signals measured at the CN Tower in the absence of lightning. A modified spectral substraction method (MSS) is proposed and applied to the de- noise the simulated current derivative signal and the resultant waveform is compared with the Heidler derivative model, which enabled accurate evaluation of the proposed method. The result of the evaluation show a substantial improvement in the signal peak-to-noisepeak ratio(SPNPR) of up to 32 dB depending on the level of vthe noise signal, which is added to the Heidler derivative function. Furthermore, 95.7%-98.5% recovery of the peak of the original Heidler derivative function was obtained. For further evaluation of the new MSS method, the conventional spectral subtraction (SS) method is applied for de-noising the same simulated current derivative signals, which produced a substantially lower SPNPR of up to 16 dB with a peak recovery of 93.3%- 97.5% of the original Heidler derivative model. The poposed method is successfully used to substantially remove the noise from the lightning current derivative signals measured at the CN Tower.

show abstract

A spectral subtraction method for the enhancement of speech corrupted by nonwhite, nonstationary noise

Cited by 13 publications

References 12 publications

A Robust Time-frequency Decomposition Model for Suppression of Mixed Gaussian-impulse Noise in Audio Signals

A Robust Time-frequency Decomposition Model for Suppression of Mixed Gaussian-impulse Noise in Audio Signals

Quantifying Human Indoor Activity Using a Software Radio-Based Radar

Enhancement of CN Tower lightning current derivative signals using a modified power spectral subtraction method

Contact Info

Product

Resources

About