Assessing local noise level estimation methods: Application to noise robust ASR

2001 ◽  
Vol 34 (1-2) ◽  
pp. 141-158 ◽  
Author(s):  
Christophe Ris ◽  
Stéphane Dupont
2019 ◽  
Vol 2019 ◽  
pp. 1-12
Author(s):  
Hanlin Tan ◽  
Huaxin Xiao ◽  
Shiming Lai ◽  
Yu Liu ◽  
Maojun Zhang

In traditional image denoising, noise level is an important scalar parameter which decides how much the input noisy image should be smoothed. Existing noise estimation methods often assume that the noise level is constant at every pixel. However, real-world noise is signal dependent, or the noise level is not constant over the whole image. In this paper, we attempt to estimate the precise and pixelwise noise level instead of a simple global scalar. To the best of our knowledge, this is the first work on the problem. Particularly, we propose a deep convolutional neural network named “deep residual noise estimator” (DRNE) for pixelwise noise-level estimation. We carefully design the architecture of the DRNE, which consists of a stack of customized residual blocks without any pooling or interpolation operation. The proposed DRNE formulates the process of noise estimation as pixel-to-pixel prediction. The experimental results show that the DRNE can achieve better performance on nonhomogeneous noise estimation than state-of-the-art methods. In addition, the DRNE can bring denoising performance gains in removing signal-dependent Gaussian noise when working with recent deep learning denoising methods.


2012 ◽  
Vol 22 (03) ◽  
pp. 1250052 ◽  
Author(s):  
PENGCHENG XU ◽  
W. K. LI ◽  
A. W. JAYAWARDENA

In this study, the correlation sum and the correlation integral for chaotic time series using the Supremum norm and the Euclidean norm are discussed. The correlation integrals are then used to develop governing equations for the correlation sum, noise level and correlation dimension in which the correlation dimension and the noise level are linearly dependent on each other. Some linear estimation methods for the noise level are then introduced by using these equations. The estimation methods are applied to four chaotic time series (two artificial and two real-world). By comparing the performances of the estimations of the noise level, the best estimating method is then suggested.


Author(s):  
Ismail El Ouargui ◽  
Said Safi ◽  
Miloud Frikel

The resolution of a Direction of Arrival (DOA) estimation algorithm is determined based on its capability to resolve two closely spaced signals. In this paper, authors present and discuss the minimum number of array elements needed for the resolution of nearby sources in several DOA estimation methods. In the real world, the informative signals are corrupted by Additive White Gaussian Noise (AWGN). Thus, a higher signal-to-noise ratio (SNR) offers a better resolution. Therefore, we show the performance of each method by applying the algorithms in different noise level environments.


2021 ◽  
Vol 13 (1) ◽  
Author(s):  
Jailos Mrisho Nzumile ◽  

Autoregressive (AR2) technique has always been used to estimate frequency of the output signal from Large ring laser. However, the acquisition rate is not at near real time which is the requirement and noise level still challenge the process resulting to errors in the final estimation. A research was done to compare the Autoregressive (AR2) with the counterparts such as Pisarenko, Quinn, Hilbert and Phase looking for a better technique that will estimate the frequency at near real time to minimize errors. Secondary data from G and C – II ring laser were used during the comparison between the techniques and Autoregressive (AR2). Results shows that, the output characteristics from the counterpart does not depict the oscillations of the Earth rotation as expected contrast to that of Autoregressive (AR2) which does. Moreover, there were much deviation from the expected true value for the techniques contrast to that of AR2 which is very minimum. On the other hand, when the C – II data were used, it was observed that both techniques resemble on their output characteristics though AR2 was still better in the acquisition rate expect for Hilbert transform which does not resemble with others. Following the scope of this paper, Autoregressive (AR2) technique still emerge as a favorite frequency estimation technique contrast to the four counterparts due to its robustness, high acquisition rate as well as low noise level.


2017 ◽  
Vol 25 (6) ◽  
pp. 907-926 ◽  
Author(s):  
Ti Bai ◽  
Hao Yan ◽  
Luo Ouyang ◽  
David Staub ◽  
Jing Wang ◽  
...  

Sign in / Sign up

Export Citation Format

Share Document