9 results
Search Results
2. A variational EM approach to predictive uncertainty
- Author
-
Harva, Markus
- Subjects
- *
HUMAN error , *REGRESSION analysis , *PAPER , *NOISE - Abstract
Abstract: In many applications of regression, the conditional average of the target variable is not sufficient for prediction. The dependencies between the explanatory variables and the target variable can be complex calling for modelling of the full conditional probability density. The ubiquitous problem with such methods is overfitting since due to the flexibility of the model the likelihood of any datapoint can be made arbitrarily large. In this paper a method for predicting uncertainty by modelling the conditional density is presented based on conditioning the scale parameter of the noise process on the explanatory variables. The model is constructed in such a manner that the unpredictability of the scale of the target distribution translates into a more robust predictive distribution. The overfitting problems are solved by learning the model using variational EM. The method is experimentally demonstrated with synthetic data as well as with real-world environmental data. The viability of the approach was put to test in the ‘Predictive uncertainty in environmental modelling’ competition held at WCCI’06. The proposed method won the competition. [Copyright &y& Elsevier]
- Published
- 2007
- Full Text
- View/download PDF
3. Nonlinear dynamic filtering of logarithmically amplified fringe signals in optical coherence tomography applied to paper measurements.
- Author
-
Alarousu, E., Bilyk, V., Gurov, I., Hast, J., Krehut, L., Myllylä, R., and Zakharov, A.
- Subjects
- *
KALMAN filtering , *PAPER , *OPTICAL coherence tomography , *LOGARITHMS , *NOISE , *SIGNAL processing - Abstract
Application of the nonlinear Kalman filtering method to logarithmically amplified low-coherence fringe signals measured from paper samples is considered. Experimental results of dynamic fringe envelope recovery in optical coherence tomography (OCT) systems are presented. The analog fringe envelope and digital dynamic fringe envelope recoveries from noisy signals are compared. The results show that the nonlinear discrete Kalman filtering method can be applied to estimate envelopes of logarithmically transformed low-coherence fringe signals with high noise immunity. Logarithmic amplification reduces quantization error in dealing with small signal values. The experimental results obtained demonstrate the possibility of purely digital signal processing in OCT. [ABSTRACT FROM AUTHOR]
- Published
- 2006
- Full Text
- View/download PDF
4. Fast detection and impulsive noise removal in color images
- Author
-
Smolka, Bogdan and Chydzinski, Andrzej
- Subjects
- *
PAPER , *NOISE , *CONSONANCE (Music theory) , *DISSONANCE (Music theory) - Abstract
Abstract: In this paper, a novel approach to the impulsive noise removal in color images is presented. The proposed technique employs the switching scheme based on the impulse detection mechanism using the so-called peer group concept. Compared to the vector median filter and other commonly used multichannel filters, the proposed technique consistently yields very good results in suppressing both the random and fixed-valued impulsive noise. The main advantage of the proposed noise detection framework is its enormous computational speed, which enables efficient filtering of color images in real-time applications. [Copyright &y& Elsevier]
- Published
- 2005
- Full Text
- View/download PDF
5. Bias minimizing filter design for gradient-based image registration
- Author
-
Robinson, Dirk and Milanfar, Peyman
- Subjects
- *
METHODOLOGY , *NOISE , *PAPER , *ETHNOMETHODOLOGY - Abstract
Abstract: Gradient-based image registration techniques represent a very popular class of approaches to registering pairs or sets of images. As the name suggests, these methods rely on image gradients to perform the task of registration. Very often, little attention is paid to the filters used to estimate image gradients. In this paper, we explore the relationship between such gradient filters and their effect on overall estimation performance in registering translated images. We propose a methodology for designing filters based on image content that minimize the estimator bias inherent to gradient-based image registration. We show that minimizing such bias improves the overall estimator performance in terms of mean square error (MSE) for high signal-to-noise ratio (SNR) scenarios. Finally, we propose a technique for designing such optimal gradient filters in the context of iterative multiscale image registration and verify their further improved performance. [Copyright &y& Elsevier]
- Published
- 2005
- Full Text
- View/download PDF
6. Analysis of long crack lines in paper webs.
- Author
-
Salminen, L. I., Alava, M. J., and Niskanen, K. J.
- Subjects
- *
FRACTURE mechanics , *PAPER , *SURFACE roughness , *CORROSION & anti-corrosives , *MATERIAL fatigue - Abstract
: We analyze 6500 mm long fracture lines of paper as an example of crack propagation involving disorder. The cracks are asymptotically self-affine, with a roughness exponent close to 0.6. Systematic deviations from the power-law-scaling exist below a lengthscale related to the microscopic heterogeneities and possibly to a cross-over from 3d to 2d crack propagation. Several analysis methods are discussed, including first return analysis and the detection of correlated trends. [ABSTRACT FROM AUTHOR]
- Published
- 2003
- Full Text
- View/download PDF
7. Run and Random--Walk for Binary Frequency Determination of Continuous Sinusoid.
- Author
-
Inomoto, Kenji, Noro, Yuichi, and Hattori, Shozo
- Subjects
- *
PAPER , *SIGNAL detection , *NOISE , *SIMULATION methods & models , *EVALUATION - Abstract
This paper considers the determination of the frequency of a sinusoidal signal from its sampled values. A method is described by which the effect of the noise is reduced and the frequency is determined accurately. The method forms the sample set composed of four samples from the signal with unknown frequency, and the binary representation of the estimation is determined successively by varying the sampling frequency. However, the noise in the signal may produce an error. One means to eliminate the effect of the noise is to determine one bit of the binary number based on several sample sets. This paper proposes the following method to reduce the estimation error for the frequency due to the noise. The binary number is partitioned into the upper and lower bits. The upper bits are determined by the property of the rum, and the lower bits are determined by the property of the ramdom-walk. In either case, each bit is determined by several sample sets. By the evaluation based on the bit error rate, the accuracy is improved by combining the forementiomed methods, compared with the majority decision. A comprehensive evaluation was also made by simulation, indicating that the r.m.s. error can be reduced to 2 percent for the signal-to-noise ratio of 20 dB. [ABSTRACT FROM AUTHOR]
- Published
- 1989
8. A History of the Human Voice.
- Author
-
Evenson, Brian
- Subjects
- *
HUMAN voice , *NOISE , *IMAGE reconstruction , *PHOTOGRAPHS , *PAPER - Abstract
The article suggests an explanation for the rediscovered noises based on a conjunction between the damaged paper and the damaged voice that appears in the technical work of image remaking. It shows how the pictorial image can slide into the aural. It proposes that while people may be damaged by the images, they also actively participate, at the moment of their creation, in the damaging of images.
- Published
- 2010
- Full Text
- View/download PDF
9. The Effect of Noise on Public Health.
- Author
-
Finegold, Lawrence S., Job, Soames, de Jong, Ronald, and Griefahn, Barbara
- Subjects
- *
NOISE , *CONFERENCES & conventions , *ANIMALS , *PAPER , *RESEARCH , *SOUND - Abstract
Focuses on the effect of noise exposure on humans and other animals during the 8th International Congress on Noise. Papers regarding sleep-disturbance research; Community response to environmental noise; Important negative effects of noise.
- Published
- 2004
Discovery Service for Jio Institute Digital Library
For full access to our library's resources, please sign in.