Showing papers by "Jie Yang published in 2006"
TL;DR: In this paper, based on the concept of representing protein samples in terms of their pseudo-amino acid composition, the fuzzy K-nearest neighbors (KNN) algorithm has been introduced to predict membrane protein types, and high success rates were observed.
167 citations
TL;DR: A novel approach called "stacked generalization" or "stacking" has been introduced, which can combine several different types of classifiers through a meta-classifier to maximize the generalization accuracy.
161 citations
14 Jun 2006
TL;DR: This paper presents a new weighted fuzzy kernel-clustering algorithm (WFKCA), which can yield the meaningful prototypes (cluster centers) of the clusters in a kernel feature space mapped by mercer kernels.
Abstract: Clustering analysis is an important topic in artificial intelligence, data mining and pattern recognition research. Conventional clustering algorithms, for instance, the famous Fuzzy C-means clustering algorithm (FCM), assume that all the attributes are equally relevant to all the clusters. However in most domains, especially for high-dimensional dataset, some attributes are irrelevant, and some relevant ones are less important than others with respect to a specific class. In this paper, such imbalances between the attributes are considered and a new weighted fuzzy kernel-clustering algorithm (WFKCA) is presented. WFKCA performs clustering in a kernel feature space mapped by mercer kernels. Compared with the conventional hard kernel-clustering algorithm, WFKCA can yield the meaningful prototypes (cluster centers) of the clusters. Numerical convergence properties of WFKCA are also discussed. For in-depth studies, WFKCA is extended to WFKCA2, which has been demonstrated as a useful tool for clustering incomplete data. Numerical examples demonstrate the effectiveness of the new WFKCA algorithm
113 citations
TL;DR: The rough set rule-based method applied to predict the degree of malignancy can achieve higher classification accuracy than other intelligent analysis methods such as neural networks, decision trees and a fuzzy rule extraction algorithm based on Fuzzy Min-Max Neural Networks.
95 citations
TL;DR: Support vector machines with floating search method is utilized to select relevant features and to predict the degree of malignancy in brain glioma and results show that the feature subset selected by the techniques can yield better classification performance.
75 citations
TL;DR: By practical experiments, it is verified that the proposed novel algorithm obtains better results than original methods, especially with respect to images within holes, complex background, weak edges, and noise.
53 citations
01 Jan 2006
TL;DR: This paper analyzes the distributing characteristic of hyper-parameters to propose a hybrid method that combines evolution strategies (ES) with a grid search (GS) to carry out optimizing selection of these hyperparameters.
Abstract: In real-world applications, selecting the appropriate hyper-parameters for support vector machines (SVM) is a difficult and vital step which impacts the generalization capability and classification performance of classifier. In this paper, we analyze the distributing characteristic of hyper-parameters that the optimal hyper-parameters points form neighborhoods. For finding all the optimal points (on the grid points) in neighborhoods, based on this characteristic, we propose a hybrid method that combines evolution strategies (ES) with a grid search (GS), to carry out optimizing selection of these hyperparameters. We firstly use evolution strategies find the optimal points of hyperparameters and secondly execute a grid search in the neighborhood of these points. Our hybrid method takes advantage of the high computing efficiency of ES and the exhaustive searching merit of GS. Experiments show our hybrid method can successfully find the optimal hyper-parameters points in neighborhoods.
44 citations
Journal Article•
TL;DR: Zhang et al. as mentioned in this paper proposed a Gabor-based supervised locality preserving projection (GSLPP) method for face recognition using class labels of data points to enhance its discriminant power in their mapping into a low dimensional space.
Abstract: This paper introduces a novel Gabor-based supervised locality preserving projection (GSLPP) method for face recognition. Locality preserving projection (LPP) is a recently proposed method for unsupervised linear dimensionality reduction. LPP seeks to preserve the local structure which is usually more significant than the global structure preserved by principal component analysis (PCA) and linear discriminant analysis (LDA). In this paper, we investigate its extension, called supervised locality preserving projection (SLPP), using class labels of data points to enhance its discriminant power in their mapping into a low dimensional space. The GSLPP method, which is robust to variations of illumination and facial expression, applies the SLPP to an augmented Gabor feature vector derived from the Gabor wavelet representation of face images. We performed comparative experiments of various face recognition schemes, including the proposed GSLPP method, principal component analysis (PCA) method, linear discriminant analysis (LDA) method, locality preserving projection method, the combination of Gabor and PCA method (GPCA) and the combination of Gabor and LDA method (GLDA). Experimental results on AR database and CMU PIE database show superior of the novel GSLPP method.
36 citations
TL;DR: A new color-texture image segmentation method that combined directional operators and the JSEG (J measure based SEGmentation) algorithm was presented and it was found the over-segmentation problem of JSEg could be attributed partly to the absence of color discontinuity in the J measure.
33 citations
TL;DR: Experimental results show that the local variance WIE filter can evidently increase the signal-to-noise ratio of preprocessed images.
Abstract: As a robust criterion to evaluate the complex degree of infrared (IR) backgrounds, the variance weighted information entropy (WIE) is applied to preprocess the IR small target images and detect the sea-sky line in sea-sky IR images. Experimental results show that the local variance WIE filter can evidently increase the signal-to-noise ratio of preprocessed images. In addition, the validity of the variance WIE based sea-sky line detection method is verified by experiments.
33 citations
TL;DR: This work presents a novel DRR generation technique, called the adaptive Monte Carlo volume rendering (AMCVR) algorithm, based on the conventional Monte CarloVolume rendering (MCVR) technique that is very efficient for rendering large medical datasets.
Abstract: Digitally reconstructed radiograph (DRR) generation is an important step in several medical imaging applications such as 2D-3D image registration, where the generation of DRR is a rate-limiting step. We present a novel DRR generation technique, called the adaptive Monte Carlo volume rendering (AMCVR) algorithm. It is based on the conventional Monte Carlo volume rendering (MCVR) technique that is very efficient for rendering large medical datasets. In contrast to the MCVR, the AMCVR does not produce sample points by sampling directly in the entire volume domain. Instead, it adaptively divides the entire volume domain into sub-domains using importance separation and then performs sampling in these sub-domains. As a result, the AMCVR produces almost the same image quality as that obtained with the MCVR while only using half samples, and increases projection speed by a factor of 2. Moreover, the AMCVR is suitable for fast memory addressing, which further improves processing speed. Independent of the size of medical datasets, the AMCVR allows for achieving a frame rate of about 15 Hz on a 2.8 GHz Pentium 4 PC while generating reasonably good quality DRR.
22 Nov 2006
TL;DR: This paper adopts an enhanced cascaded tree style learner framework for car plate detection using the hybrid object features including the simple statistical features and Harr-like features to reduce the false alarm and the false dismissal while retaining a high detection ratio.
Abstract: Car plate detection is a key component in automatic license plate recognition system. This paper adopts an enhanced cascaded tree style learner framework for car plate detection using the hybrid object features including the simple statistical features and Harr-like features. The statistical features are useful for simplifying the process on cascade classifier. The cascaded tree-style detector design will further reduce the false alarm and the false dismissal while retaining a high detection ratio. The experimental results obtained by the proposed algorithm exhibit the encouraging performance.
18 Sep 2006
TL;DR: A novel Gabor-based supervised locality preserving projection (GSLPP) method for face recognition is introduced, using class labels of data points to enhance its discriminant power in their mapping into a low dimensional space.
Abstract: This paper introduces a novel Gabor-based supervised locality preserving projection (GSLPP) method for face recognition. Locality preserving projection (LPP) is a recently proposed method for unsupervised linear dimensionality reduction. LPP seeks to preserve the local structure which is usually more significant than the global structure preserved by principal component analysis (PCA) and linear discriminant analysis (LDA). In this paper, we investigate its extension, called supervised locality preserving projection (SLPP), using class labels of data points to enhance its discriminant power in their mapping into a low dimensional space. The GSLPP method, which is robust to variations of illumination and facial expression, applies the SLPP to an augmented Gabor feature vector derived from the Gabor wavelet representation of face images. We performed comparative experiments of various face recognition schemes, including the proposed GSLPP method, principal component analysis (PCA) method, linear discriminant analysis (LDA) method, locality preserving projection method, the combination of Gabor and PCA method (GPCA) and the combination of Gabor and LDA method (GLDA). Experimental results on AR database and CMU PIE database show superior of the novel GSLPP method.
TL;DR: The proposed hybrid model-based method for obtaining an accurate and topologically-preserving segmentation of the brain cortex is based on defining region and boundary information using, respectively, level set and Bayesian techniques, and fusing these two types of information to achieve cerebral cortex segmentation.
TL;DR: An improved approach for J value segmentation (JSEG) is presented for unsupervised color-texture segmentation and shows that the improved method overcomes the limitations of JSEG successfully and is more robust.
TL;DR: This paper investigates its extension, called supervised locality pursuit embedding (SLPE), using class labels of data points to enhance its discriminant power in their mapping into a low dimensional space and demonstrates that SLPE is superior to other three methods in terms of recognition accuracy.
Journal Article•
TL;DR: Face Detection and Recognition using Colour Sequential Images Zhonglong Zheng Institute of Information Science and Engineering, Zhejiang Normal University, Jinhua, 321004,Zhejiang, China E-mail: zhonglong@sjtu.edu.cn
Abstract: A human face detection and recognition system for colour sequential images is presented in this paper. The system is composed of two subsystems: human face detection subsystem, human face recognition subsystem. The face detection subsystem includes two modules: face finding, and face verification. The human face finding module finds the candidate face regions from colour sequential images according to skin colour analysis and motion analysis. The human face verification module has been developed to verify the detected human faces by judging of eclipse and Support Vector Machines (SVM), and precisely localize human faces by locating eyes and mouths based on Generalized Symmetry Transform. The features of the relation between face patterns can be extracted and selected by Principal Component Analysis. The selected features are used to train multiple SVMs which can finally classify human faces. The system structure is designed according to the following principle: Firstly simpler methods are used to reduce the search space, and then more complicated methods are used in the reduced space. So the system can have a quick response speed as well as holding high detection and recognition rate. Human face detection accuracy of the system is 97.2% under controllable lightning condition. Human face (70 persons) recognition accuracy of the system is 96.5% (with 20 eigenvectors) and 98.3% (with 30 eigenvectors).
TL;DR: A distortion-tolerant reconstruction algorithm for sweep fingerprint sequences oriented to portable devices is proposed and the performance of this algorithm on simulated and real data is presented.
Abstract: A distortion-tolerant reconstruction algorithm for sweep fingerprint sequences oriented to portable devices is proposed. The performance of this algorithm on simulated and real data is presented. Some performance results are included to illustrate the efficiency of this algorithm compared to two state-of-the-art reconstruction methods.
TL;DR: Robust face detection is implemented firstly using probabilistic method, and the chin contours are extracted accurately using the active shape model (ASM), which depends on the parameters obtained from the face detection.
Patent•
19 Jul 2006
TL;DR: In this article, the authors proposed a method for extracting the tongue tip character area, the root character area and the seed points around the tongue body using a color gradient image. But the method is not suitable for the extraction of the entire tongue.
Abstract: The invention relates to a tongue body extracting method in the field of image processing technology. It comprises: ascertaining the tongue tip character area, ascertaining the tongue root character area, generating color gradient picture, constructing Snake initial curve, cutting initially, quoting the cutting effect: directly extracting the tongue body image when it cuts accurately, or enters into the next step; ascertaining some seed points around the tongue body, auto generating Pit line by the color gradient picture and doing second cutting: the corresponding Snake curve has been equally divided, the end point of each section is connected with the corresponding point of the Pit line, computing the spring force between the Pit points, adding the external force in the energy function, iterating the Snake curve until it contracts, at last extracting the tongue body image.
TL;DR: An efficient volume rendering method is proposed by dividing the volume rendering integral into four sub-Integrals and enabling sampling in each sub-integral to be "best" while achieving viewing independency, and the results show that thus rendered images exhibit high quality.
13 Dec 2006
TL;DR: Wang et al. as discussed by the authors proposed three extended fitting methods to the standard active shape model (ASM) to improve the accuracy or speed in a way, and the combination of three methods improved the accuracy and speed greatly.
Abstract: In this study, we propose three extended fitting methods to the standard ASM(active shape model). Firstly, profiles are extended from 1D to 2D; Secondly, profiles of different landmarks are constructed individually; Thirdly, length of the profilesis determined adaptively with the change of level during searching, and the displacements in the last level are constrained. Each method and the combination of three methods are tested on the SJTU(Shanghai Jiaotong University) face database. In all cases, compared to the standard ASM, each method improves the accuracy or speed in a way, and the combination of three methods improves the accuracy and speed greatly.
21 Jul 2006
TL;DR: A novel backward tracking method is introduced to solve scaling problem, and the solution of dealing with the severe object motions is also discussed by integrating mean-shift tracker into the low-resolution matching scheme.
Abstract: The classic mean-shift tracker has no integrated scale adaptation, which limits its performance in tracking variable scale object as wel l as the object with severe motions. Based on the variation analysis of Bhattacharyya coefficient within mean-shift framework, the sufficient conditions for accurate tracking of object with scale changes are presented. We propose that the changes of object scale and position within the region of previous tracking window will not impact the localization accuracy of mean-shift tracker. Based on our findings, a novel backward tracking method is introduced to solve scaling problem, and the solution of dealing with the severe object motions is also discussed by integrating mean-shift tracker into the low-resolution matching scheme.
13 Dec 2006
TL;DR: A novel fingerprint feature named the octantal nearest-neighbor structure (ONNS) is defined and based on the ONNS, the minutiae pairing algorithm is conducted, and a novel algorithm is developed to evaluate the translational and rotational parameters between the input and the template fingerprints.
Abstract: In this paper, we propose a novel Octantal Nearest-neighbor Structure and core points based fingerprint matching scheme. A novel fingerprint feature named the octantal nearest-neighbor structure (ONNS) is defined. Based on the ONNS, the minutiae pairing algorithm is conducted to find the corresponding minutiae pairs, and a novel algorithm is developed to evaluate the translational and rotational parameters between the input and the template fingerprints. Core point based orientation pairing is performed thereafter. Matching score is calculated. Experimental results on the FVC2004 fingerprint databases show the good performance of the proposed algorithm.
TL;DR: In this article, three novel Ba5LnNiTa9O30 (Ln = La, Nd and Sm) ceramics were prepared and characterized in the BaO-Ln2O3-NiO-Ta2O5 system.
Abstract: Three novel Ba5LnNiTa9O30 (Ln = La, Nd and Sm) ceramics were prepared and characterized in the BaO-Ln2O3-NiO-Ta2O5 system. All three compounds adopted the filled tetragonal tungsten bronze (TB) structure at room temperature. The present ceramics exhibited relaxor behavior, and the Curie temperature (at 10kHz) were −130, −80 and −45°C for Ba5LaNiTa9O30, Ba5NdNiTa9O30, and Ba5SmNiTa9O30 respectively. At room temperature, Ba5LnNiTa9O30 ceramics have a high dielectric constants in the range 102∼118, a low dielectric loss in range 0.0019∼0.0036, and the temperature coefficients of the dielectric constant (τɛ) in the range −320∼−460 ppm°C−1 (at 1 MHz).
TL;DR: A coarse-to-fine image registration scheme is proposed, which first uses the block-matching method as the preregistration and then adopts the curvature-based elastic registration as a fine estimation of nonlinear distortions between two consecutive frames.
Abstract: For reconstructing sweep fingerprint sequences, we propose a coarse-to-fine image registration scheme, which first uses the block-matching method as the preregistration and then adopts the curvature-based elastic registration as a fine estimation of nonlinear distortions between two consecutive frames. With respect to two existing state-of-the-art approaches, the new scheme becomes more robust against distortions and therefore the reconstruction accuracy is improved.
TL;DR: First, face images are transformed by Gabor filters to obtain a set of overcompleted feature vectors, which can remove intrinsic redundancies within images and provide orientation-selective properties to enhance differences among face poses as well.
Abstract: We investigate the appearance manifold of differ- ent face poses using manifold learning. The pose estimation problem is, however, exacerbated by changes in illumina- tion, spatial scale, etc. In addition, manifold learning has some disadvantages. First, the discriminant ability of the low-dimensional subspaces obtained by manifold learning often is lower than traditional dimesionality reduction ap- proaches. Second, manifold learning methods fail to remove the redundancy, such as high-order correlation, among origi- nal feature vectors. In this work, we propose a novel ap- proach to address these problems. First, face images are transformed by Gabor filters to obtain a set of overcompleted feature vectors, which can remove intrinsic redundancies within images and provide orientation-selective properties to enhance differences among face poses as well. Second, su- pervised locality preserving projections SLPPs are pro- posed to reduce dimensionality and obtain the low- dimensional subspace, which has the ability to maximize between-class distance and minimize within-class distance. Finally, the support vector machine SVM classifier is ap- plied to estimate face poses. The experimental results show that the proposed approach is effective and efficient. © 2006
21 Aug 2006
TL;DR: A non-negative matrix factorisation (NMF)-based relevance feedback approach is introduced that uses a standard NMF algorithm to construct a reliable semantic space from a pool of relevant images based on a user's interactions, which results in the semantic space being closer to the user's expectation.
Abstract: As a powerful tool for content-based image retrieval, many techniques have been proposed for relevance feedback. A non-negative matrix factorisation (NMF)-based relevance feedback approach is introduced. This approach uses a standard NMF algorithm to construct a reliable semantic space from a pool of relevant images based on a user's interactions, because the latent semantic space derived by NMF does not need to be orthogonal, and each image is guaranteed to take only non-negative values in all the latent semantic directions. It means that each axis in the space derived by NMF has a straightforward correspondence with each image semantic class. In addition, the hidden semantic features of the query and images in the database are extracted with an NMF-projecting algorithm. By memorising the feedback information provided by the user, the knowledge accumulated from past relevance interaction is used to update semantic space, which results in the semantic space being closer to the user's expectation. The experiments show that the proposed NMF-based relevance feedback approach performs better than other relevance feedback approaches.
01 Aug 2006
TL;DR: This paper presents a localization method without on-body sensor (LWOS), which from the detected attenuation of Received Signal Strength Indication (RSSI), LWOS can detect and localize people directly utilizing the wireless communication in WSNs.
Abstract: In many applications of wireless sensor networks (WSNs), the location information of users is very important. In this paper we present a localization method without on-body sensor (LWOS). The basic idea is that when a person is standing between a pair of transceivers, the human body will attenuate the received signal. From the detected attenuation of Received Signal Strength Indication (RSSI), LWOS can detect and localize people directly utilizing the wireless communication in WSNs. No additional sensor is needed and users do not need to wear a sensor node any more. A signal-shielding device is used at the transmitter side to minify the interference of RSSI variability from multi-path effects. Experiment results show a good capability of localizing a single user in an indoor environment.
Journal Article•
TL;DR: Variational Bayesian method for source separation which models additive noise into the mixing system and a low order autoregressive (AR) model is adopted to describe the temporal structure of source.
Abstract: We propose to use a low order autoregressive (AR) model to describe the temporal structure of source. Then we adopt Variational Bayesian (VB) method for source separation which models additive noise into the mixing system. The approach integrates the source probabilistic model and noise probabilistic model. Its goal is to approximate the actual probability density function of hidden variables and parameters using its approximating posterior distribution by minimizing the misfit between them. The advantage of our VB-AR algorithm is that it can exploit the temporal nature of source signals and avoid over-fitting in the separating process. This algorithm can also identify the AR order. Experiments on artifact and real-world speech signals are used to verify our proposed algorithms. As a result, the lower AR source model improves the separation. The performance of the algorithm is compared with that of i.i.d. separation algorithm and the second-order separation algorithm.