A method of single channel speech enhancement is proposed by de-noising using stationary wavelet transform. The approach developed herein processes multi-resolution wavelet coefficients individually and then recovery ...
详细信息
A method of single channel speech enhancement is proposed by de-noising using stationary wavelet transform. The approach developed herein processes multi-resolution wavelet coefficients individually and then recovery signal is reconstructed. The time invariant characteristics of stationary wavelet transform is particularly useful in speech de-noising. Experimental results show that the proposed speech enhancement by de-noising algorithm is possible to achieve an excellent balance between suppresses noise effectively and preserves as many target characteristics of original signal as possible. This de-noising algorithm offers a superior performance to speech signal noise suppress.
A new scheme for approximating the solution of 2D Maxwell's equations using the symplectic scheme is introduced. The scheme is obtained by discretizing the Maxwell' s equations in the time direction based on s...
详细信息
A new scheme for approximating the solution of 2D Maxwell's equations using the symplectic scheme is introduced. The scheme is obtained by discretizing the Maxwell' s equations in the time direction based on symplectic scheme with different orders, and then evaluated the equation in the spatial direction with a second or fourth order finite difference approximation. The stability condition and numerical dispersion of the schemes with different orders are derived. The results are demonstrated by theoretical analysis and numerical simulation, the stability and numerical dispersion of the scheme with first and second order symplectic scheme (T1S2, T2S2) are identical to FDTD with a second order approximation in spatial direction. Although the high order schemes have almost the same stability as the FDTD, the fourth order scheme with a fourth order approximation in spatial direction (T4S4) has the superior numerical dispersion-isotropic properties of the scheme. Numerical results show that high order symplectic scheme is superior compared with FDTD for solving two-dimensional TMz case.
Label is denoted by disparity and the energy function is established. Then the problem of matching can be transformed into that of energy function minimization. A network is constructed such that the energies can be r...
详细信息
Label is denoted by disparity and the energy function is established. Then the problem of matching can be transformed into that of energy function minimization. A network is constructed such that the energies can be related to the capacities of the cuts of the network. Finally, the minimal energy is obtained by the network-flows theory, and hence the disparity data are solved. Comparing with some known algorithms based on graph cuts, the algorithm extends the label from 1 dimension vector to 2 dimension vector, and adapts vision matching of more general conditions;furthermore the algorithm can gain the minimization in global. Experimental results show that the algorithm has a high accuracy.
A novel and efficient speckle noise reduction algorithm based on Bayesian wavelet shrinkage using cycle spinning is proposed. First, the sub-band decompositions of non-logarithmically transformed SAR images are shown....
详细信息
A novel and efficient speckle noise reduction algorithm based on Bayesian wavelet shrinkage using cycle spinning is proposed. First, the sub-band decompositions of non-logarithmically transformed SAR images are shown. Then, a Bayesian wavelet shrinkage factor is applied to the decomposed data to estimate noise-free wavelet coefficients. The method is based on the Mixture Gaussian Distributed (MGD) modeling of sub-band coefficients. Finally, multi-resolution wavelet coefficients are reconstructed by wavelet-threshold using cycle spinning. Experimental results show that the proposed despeckling algorithm is possible to achieve an excellent balance between suppresses speckle effectively and preserves as many image details and sharpness as possible. The new method indicated its higher performance than the other speckle noise reduction techniques and minimizing the effect of pseudo-Gibbs phenomena.
Based on modeling idea of partial least squares (PLS) and divided the values of response variable into two classes denoted by 0 and 1, a novel classification algorithm of land quality is set up in this paper. Firstly,...
详细信息
Based on modeling idea of partial least squares (PLS) and divided the values of response variable into two classes denoted by 0 and 1, a novel classification algorithm of land quality is set up in this paper. Firstly,...
详细信息
Based on modeling idea of partial least squares (PLS) and divided the values of response variable into two classes denoted by 0 and 1, a novel classification algorithm of land quality is set up in this paper. Firstly, the algorithms of multiple linear regression (MLR) and principal component regression (PCR) are introduced and analysed their shortages. Then on the basis of modeling idea of PLS, the classification algorithm of land quality is constructed. The experiment shows that the PLS algorithm doesn't request distribution of the data, and has best classification pattern ability compared with the algorithms of MLR and PCR. It has more advantages than MLR, PCR, such as simplicity and robustness, clearly qualitative explanation. It is powerful for multicollinearity, particularly when the number of predictor variables is large and the sample size is small, and provides a novel research method for classification of land quality
This paper studies 3D surface integration from novel angle. The authors consider the construction to the multiple surface patches, not one surface, from the Gauss map. The algorithm takes as its input a 2D field of su...
详细信息
This paper studies 3D surface integration from novel angle. The authors consider the construction to the multiple surface patches, not one surface, from the Gauss map. The algorithm takes as its input a 2D field of surface normal estimates, delivered, for instance, by a shape-from-shading or shape-from-texture procedure. The authors disintegrate the Gauss map into two functions by the spherical coordinates, and then borrow the ideas from routine image processing theory to filter the two functions and to segment the space surface into several subsurface, at last the authors use the integrability to recover the subsurface individually. The method only exploits the general techniques in image processing, but can supply better results than the previous researches, which are only based on one function model, especially in the preservence the edges between different surfaces. The method is evaluated on synthetic and real data delivered by a shape-from-shading algorithm. The approach provides an actually way to use the normal maps of the surface.
Gait recognition is used to identify individuals in image sequences by the way they walk. Nearly all of the approaches proposed for gait recognition are 2D methods based on analyzing image sequences captured by a sing...
详细信息
Gait recognition is used to identify individuals in image sequences by the way they walk. Nearly all of the approaches proposed for gait recognition are 2D methods based on analyzing image sequences captured by a single camera. In this paper, video sequences captured by multiple cameras are used as input, and then a human 3D model is set up. The motion is tracked by applying a local optimization algorithm. The lengths of key segments are extracted as static parameters, and the motion trajectories of lower limbs are used as dynamic features. Finally, linear time normalization is exploited for matching and recognition. The proposed method based on 3D tracking and recognition is robust to the changes of viewpoints. Moreover, better results are achieved for sequences containing difficult surface variations than with 2D methods, which prove the efficiency of our algorithm
In this paper, we propose a method to find the anomalous behaviors in network traffic. We map the network connection records into different feature spaces typically of high dimension according to their protocols and s...
详细信息
ISBN:
(纸本)1424406048
In this paper, we propose a method to find the anomalous behaviors in network traffic. We map the network connection records into different feature spaces typically of high dimension according to their protocols and services. In training, we perform clustering to group training data points into clusters, from which we select some clusters as normal and known-attack profile according to a simple, but effect, criterion. For those training data excluded from the profile, we use them to build a specific classifier. The classifier has two distinct characteristics: one is that it regards each data point in the feature space with the limited influence scope, which is served as the decisive bounds of the classifier, and the other is that it has the "default" label to recognize those novel attacks. We present a novel classification algorithm, influence-based classification algorithm, to deal with ambiguous data. Our system is tested on the KDD Cup 1999 data. Results show that it is superior to other data mining based approaches in detection performance, especially in detection of PROBE and U2R attacks
In this paper, an information pattern recognition method based on fuzzy control is set up. On one hand, the modeling method of fuzzy information classified recognition pattern has been established. On the other hand, ...
详细信息
In this paper, an information pattern recognition method based on fuzzy control is set up. On one hand, the modeling method of fuzzy information classified recognition pattern has been established. On the other hand, the data from Qufu City of Shandong Province during 14 years from 1990 to 2003 is processed and analyzed. The average temperature (℃) and rainfall (mm) in April each year are considered as the input of the system, a number of Aphis gossypii Glover (AGG)occurred for the Cotton in high period are considered as the output, Fuzzy information classified recognition pattern is set up in order to recognize the occurrence degree of the *** results of the returning recognition from 1990 to 2003 and the recognition for 2004 are satisfactory.
暂无评论