Rough set based attribute significance measure and reduction is proposed in this paper, after we decompose textures using wavelet packet and extract the l(1) -norm as features, condition attributes are discretized wit...
详细信息
ISBN:
(纸本)9781424422388
Rough set based attribute significance measure and reduction is proposed in this paper, after we decompose textures using wavelet packet and extract the l(1) -norm as features, condition attributes are discretized with equal width binning method. We deduce the classification rules withthe selected feature subset. the classification performance is tested on a set of 13 Brodatz texture, the averaged classification results show that the proposed algorithm can get rid of redundancy and only a few of the features can fulfill the classification task without reducing accuracy.
the motor is the workhorse of industry. the control and identification of induction motor drives using artificial intelligence is the key point for high performance electrical driving. A new architecture of nonlinear ...
详细信息
ISBN:
(纸本)9781424422388
the motor is the workhorse of industry. the control and identification of induction motor drives using artificial intelligence is the key point for high performance electrical driving. A new architecture of nonlinear autoregressive moving average model based on wavelet neural networks is presented for enhancing the performance of induction motor. the Akaike's final predication error criterion is applied to select the optimum number of wavelets to be used in the WNN model. By two-phase synchronously rotating reference frame transformation, an induction motor can be controlled like a separately excited dc motor. the WNN controller is utilized as speed controller to control the torque by the quadrature axis of the stator current. the WNN controller can be trained well. theoretic analysis and simulations show that the novel method is highly effective.
In this paper, we present an improved face recognition algorithm based on adjacent pixel intensity difference quantization (APIDQ) histogram method proposed by Kotani et al. [12]. We optimize the quantization method o...
详细信息
ISBN:
(纸本)9781424422388
In this paper, we present an improved face recognition algorithm based on adjacent pixel intensity difference quantization (APIDQ) histogram method proposed by Kotani et al. [12]. We optimize the quantization method of APIDQ according to the maximum entropy principle (MEP), and determine the best parameters for APIDQ. Experimental results show maximum average recognition rate of 97.2 % for 400 images of 40 persons (10 images per person) from the publicly available AT&T face database.
Handwriting-based writer identification is a hot research filed in patternrecognition. Off-line text-independent writer identification still remains as a challenging problem because writing features can only be extra...
详细信息
ISBN:
(纸本)9781424422388
Handwriting-based writer identification is a hot research filed in patternrecognition. Off-line text-independent writer identification still remains as a challenging problem because writing features can only be extracted from the handwriting images. As a result, plenty of dynamic writing information, which is very valuable for writer identification, is unavailable for off-line writer identification. this results in high error rate in off-line writer identification. In order to enhance the performance of off-line writer identification, a complex wavelet-based GGD method was presented in this paper. the novel method is based on our discovery that complex wavelet coefficients within each high-frequency sub-band of the handwritings satisfy GGD distribution. Our experiments show the new method, compared withthe traditional wavelet-based GGD method, and our method achieves a better performance.
the framework of traditional speech enhancement algorithm based on the wavelet package transformation is composed of three parts, which lies in the decomposition to the noisy speech signal based on wavelet package tra...
详细信息
ISBN:
(纸本)9781424422388
the framework of traditional speech enhancement algorithm based on the wavelet package transformation is composed of three parts, which lies in the decomposition to the noisy speech signal based on wavelet package transformation (WPD), the denoising to the coefficients of wavelet decomposition and the reconstruction of the clean speech signal. the proposed algorithm focuses on three points to improve the performance of the traditional speech enhancement algorithm. In the first part, special decomposition of the wavelet package is adopted to mimic the model of the human ears (WPDHE). In the second part, the adaptive thresh estimation (AthRSH) and the combination of soft threshold and the modified hard threshold functions (CSMHDF) are used. the experiments prove that this algorithm is more sufficient than the traditional speech enhancement algorithm.
Independent Component analysis (ICA) is a useful method for blind source separation of two signals or more. We have previously proposed a new method combining ICA withthe complex discrete wavelet transform (CDWT). In...
详细信息
ISBN:
(纸本)9781424422388
Independent Component analysis (ICA) is a useful method for blind source separation of two signals or more. We have previously proposed a new method combining ICA withthe complex discrete wavelet transform (CDWT). In this case, the voice and the noise were separated using a new method. At that time, we used the simulation signal. In this study, we analyze measured biological signals by using this new method, and discuss its effectiveness. As an example, we tried the separation of the EMG signal and the ECG signal.
Based on Canny's edge detecting criteria, we study two types of the cardinal B-spline wavelets which converge asymptotically to canny operator and Marr-Hildreth operator respectively, derive the mathematical formu...
详细信息
ISBN:
(纸本)9781424422388
Based on Canny's edge detecting criteria, we study two types of the cardinal B-spline wavelets which converge asymptotically to canny operator and Marr-Hildreth operator respectively, derive the mathematical formulas of the signal-to-noise ratio SNR, the localization L and the nonedge-mean-distance M, and prove that when detecting the object edge, the B-spline wavelet converging to Canny operator is superior to the B-spline wavelet converging to Marr-Hildreth operator. Like the first-derivative of Gauss function, B-spline wavelet converging to canny operator can satisfy the optimal criteria approximately and detect edges with very high calculation speed.
the curvelet transform is a new multi-scale analysis method which can be used to represent perfectly the curve edges of natural image. Robust image watermark algorithm should embed the watermark into the edges of imag...
详细信息
ISBN:
(纸本)9781424422388
the curvelet transform is a new multi-scale analysis method which can be used to represent perfectly the curve edges of natural image. Robust image watermark algorithm should embed the watermark into the edges of image according to Human Visual System. In this paper, an adaptive image watermark algorithm in curvelet domain was proposed. Firstly, the original image was decomposed into curvelet coefficients using fast curvelet transform. then the significant coefficient's position of different scales and orientations was located by analyzing the properties of them. At last, an adaptive watermark algorithm was used to embed a binary image into the curvelet significant coefficients by using the different scaling factor. the experiments show that a trade-off between the robust and imperceptibility can be obtained.
this paper presents a technique for speckle reduction in SAR images by using the interscale multiplication in Mallat wavelet transform (MWT) and stationary wavelet transform (SWT). the edge and non-edge regions can be...
详细信息
ISBN:
(纸本)9781424422388
this paper presents a technique for speckle reduction in SAR images by using the interscale multiplication in Mallat wavelet transform (MWT) and stationary wavelet transform (SWT). the edge and non-edge regions can be detected from median of squared amplitude in each scale. Applying this technique, the large wavelet coefficients generated by edge region and the small wavelet coefficients or speckle noises are then suppressed by using soft thresholding in all high frequency hands. the well-known threshold estimation for soft thresholding based on SimpleShrink, NormalShrink, VisuShrink, SureShrink, and BayesShrink. the despeckled image is then obtained by reconstruction from the resulted coefficients. the assessment of the proposed method, the experiment has conducted using SAR images from JERS-1 satellite. the experimental result shows that the proposed method yields noise suppression and preserves the detail feature of the images, as well as perceptual image quality.
Face representation using Gabor features and discrete wavelet has attracted considerable attention in computer vision;image processing, patternrecognition and human machine interaction. In this paper, Gabor and Haar ...
详细信息
ISBN:
(纸本)9781479959341
Face representation using Gabor features and discrete wavelet has attracted considerable attention in computer vision;image processing, patternrecognition and human machine interaction. In this paper, Gabor and Haar wavelet based feature extraction methods are proposed for the extraction of features from facial images. Face recognition experiments were carried out by using Artificial Neural Networks like MLP and RBF classifier. We've used two facial databases the ORL and computer vision. Good recognition rate were obtained using MLP classifier.
暂无评论