Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344010
Zhimao Lu, Hongxia Yu, Dongmei Fan, Chaoyue Yuan
In this paper, methods of feature selection used in the spam filtering are studied, including CHI square (CHI), Expected Cross Entropy (ECE), the Weight of Evidence for Text (WET) and Information Gain (IG) and a novel modified CHI feature selection method is proposed in spam filtering. The spam filter combined Support Vector Machine (SVM) is selected to evaluate the CHI square, Expected Cross Entropy, the Weight of Evidence for Text, Information Gain and modified CHI. The experiment proved that the modified CHI could improve the precision, recall and F test measure of spam filter and the modified CHI feature selection method is effective.
{"title":"Spam Filtering Based on Improved CHI Feature Selection Method","authors":"Zhimao Lu, Hongxia Yu, Dongmei Fan, Chaoyue Yuan","doi":"10.1109/CCPR.2009.5344010","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344010","url":null,"abstract":"In this paper, methods of feature selection used in the spam filtering are studied, including CHI square (CHI), Expected Cross Entropy (ECE), the Weight of Evidence for Text (WET) and Information Gain (IG) and a novel modified CHI feature selection method is proposed in spam filtering. The spam filter combined Support Vector Machine (SVM) is selected to evaluate the CHI square, Expected Cross Entropy, the Weight of Evidence for Text, Information Gain and modified CHI. The experiment proved that the modified CHI could improve the precision, recall and F test measure of spam filter and the modified CHI feature selection method is effective.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117305320","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Fourier transform is a widely used image processing technology. Kernel discriminant analysis is an effective nonlinear feature extraction technique. Based on them, we propose a novel feature extraction approach for face recognition. First, we perform the Fourier transform on face images and express the Fourier frequency bands in the plural form. By computing the kernel-plural discriminant capability of every frequency band, we choose the bands with strong capabilities and use them to form a new sample set. Then, we extract nonlinear discriminant features from the set and classify it by using the nearest neighbor classifier. Experimental results on AR and Feret face databases demonstrate the effectiveness of the proposed approach.
{"title":"Kernel-Plural Discriminant Analysis Based on Fourier Transform and Its Application to Face Recognition","authors":"Sheng Li, Xiaoyuan Jing, Qian Liu, Yanyan Lv, Yong-Fang Yao, Wenying Ma, Wei Xu","doi":"10.1109/CCPR.2009.5344052","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344052","url":null,"abstract":"Fourier transform is a widely used image processing technology. Kernel discriminant analysis is an effective nonlinear feature extraction technique. Based on them, we propose a novel feature extraction approach for face recognition. First, we perform the Fourier transform on face images and express the Fourier frequency bands in the plural form. By computing the kernel-plural discriminant capability of every frequency band, we choose the bands with strong capabilities and use them to form a new sample set. Then, we extract nonlinear discriminant features from the set and classify it by using the nearest neighbor classifier. Experimental results on AR and Feret face databases demonstrate the effectiveness of the proposed approach.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131549783","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344016
Hui Ding, Zhenmin Tang, Yanping Li
In the field of speaker recognition, the Gaussian Mixture Model with diagonal covariance matrices is a popular technique, in this way, it simplified model and reduced the amount of computation, but lost the correlation information between feature vectors, and then influenced the classification performance. In this paper, in order to compensate the correlation between feature elements, we proposed a novel method based on clustering transformation algorithm, we calculate the similarity between Gaussian components, and the cluster of same components will share one transformation matrix, thus multi-transformation matrices, together with weights and means vectors are obtained simultaneously by Maximum Likelihood estimation. Theory analysis and experimental results demonstrated that this proposed method can get a better balance between training speed and recognition rate, improve the performance of classifier and reduce the complexity and memory burden relatively.
{"title":"Research on the Parameter Optimal Algorithm of Gaussian Mixture Model in Speaker Identification","authors":"Hui Ding, Zhenmin Tang, Yanping Li","doi":"10.1109/CCPR.2009.5344016","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344016","url":null,"abstract":"In the field of speaker recognition, the Gaussian Mixture Model with diagonal covariance matrices is a popular technique, in this way, it simplified model and reduced the amount of computation, but lost the correlation information between feature vectors, and then influenced the classification performance. In this paper, in order to compensate the correlation between feature elements, we proposed a novel method based on clustering transformation algorithm, we calculate the similarity between Gaussian components, and the cluster of same components will share one transformation matrix, thus multi-transformation matrices, together with weights and means vectors are obtained simultaneously by Maximum Likelihood estimation. Theory analysis and experimental results demonstrated that this proposed method can get a better balance between training speed and recognition rate, improve the performance of classifier and reduce the complexity and memory burden relatively.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130919380","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344103
P. Xiao, G. Zhao, You-ping Chen
The ellipse detection is one of the classic problems in digital image processing, and has broad application prospects in machine vision, especially in automatic detection and assembly. In this paper, a novel algorithm for ellipse detection based on geometry features is presented. With the help of projection methods, the center of ellipse is detected and used to reduce the dimension of ellipse parameter space. The algorithm is applied on a set of synthetic images and industrial images in simulation and results show that the algorithm has achieved desirable detection performance.
{"title":"An Algorithm for Ellipse Detection Based on Geometry","authors":"P. Xiao, G. Zhao, You-ping Chen","doi":"10.1109/CCPR.2009.5344103","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344103","url":null,"abstract":"The ellipse detection is one of the classic problems in digital image processing, and has broad application prospects in machine vision, especially in automatic detection and assembly. In this paper, a novel algorithm for ellipse detection based on geometry features is presented. With the help of projection methods, the center of ellipse is detected and used to reduce the dimension of ellipse parameter space. The algorithm is applied on a set of synthetic images and industrial images in simulation and results show that the algorithm has achieved desirable detection performance.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"460 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114103674","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
For the purpose of color image segmentation, an unsupervised peak value searching algorithm was proposed, which was used to determine the approximate dominant color components of image. First, the local peaks of 3D color histogram within the neighborhood of 3×3 ×3 were located. The corresponding color values of local peaks were regarded as initial clustering centers, and the number of local peaks were taken as the number of clustering. In addition, taking into account of the color difference induced by local illumination, the feature vector was constructed including color and texture features. Finally, K-means clustering algorithm was applied to segment the color image. Experiment results show that the proposed method can segment the color image accurately, corresponding with the human visual. Clustering number was determined adaptively, and the problem of over-segmentation was solved effectively. The segmentation result was benefit for the following steps in the computer vision.
{"title":"Color Image Segmentation Using Combined Information of Color and Texture","authors":"Fengling Zhang, Guili Xu, Yong Zhang, Yuehua Cheng, Jingdong Wang, Yupeng Tian","doi":"10.1109/CCPR.2009.5344104","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344104","url":null,"abstract":"For the purpose of color image segmentation, an unsupervised peak value searching algorithm was proposed, which was used to determine the approximate dominant color components of image. First, the local peaks of 3D color histogram within the neighborhood of 3×3 ×3 were located. The corresponding color values of local peaks were regarded as initial clustering centers, and the number of local peaks were taken as the number of clustering. In addition, taking into account of the color difference induced by local illumination, the feature vector was constructed including color and texture features. Finally, K-means clustering algorithm was applied to segment the color image. Experiment results show that the proposed method can segment the color image accurately, corresponding with the human visual. Clustering number was determined adaptively, and the problem of over-segmentation was solved effectively. The segmentation result was benefit for the following steps in the computer vision.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114166870","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344153
L. Hong, Zhi-cheng Ji, C. Gong
Particle swarm optimization (PSO) has poor diversity, slow convergence speed and is easy to trap into local optimum during the course of searching, a modified particle swarm optimization algorithm based on immune mechanism is proposed. The new algorithm has both the properties of the original particle swarm optimization algorithm and the immune diversity keeping mechanism, and can improve the abilities of seeking the global optimum and evolution speed. The simulation results of multi-modal function optimization show that the proposed algorithm can inhibit premature effectively and has preferable global convergent performance.
{"title":"Study on Immune PSO Hybrid Optimization Algorithm","authors":"L. Hong, Zhi-cheng Ji, C. Gong","doi":"10.1109/CCPR.2009.5344153","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344153","url":null,"abstract":"Particle swarm optimization (PSO) has poor diversity, slow convergence speed and is easy to trap into local optimum during the course of searching, a modified particle swarm optimization algorithm based on immune mechanism is proposed. The new algorithm has both the properties of the original particle swarm optimization algorithm and the immune diversity keeping mechanism, and can improve the abilities of seeking the global optimum and evolution speed. The simulation results of multi-modal function optimization show that the proposed algorithm can inhibit premature effectively and has preferable global convergent performance.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114487303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344157
Lei Gu, F. Sun
Semi-supervised clustering takes advantage of a small amount of labeled data to bring a great benefit to the clustering of unlabeled data. Based on a novel kernel method for clustering using one-class support vector machine, this paper presents two novel kernel-based semi-supervised clustering methods inspired by two semi-supervised variants of the k-means clustering algorithm by seeding respectively. To investigate the effectiveness of our approaches, experiments are done on three real datasets. Experimental results show that the proposed methods can improve the clustering performance significantly compared to other unsupervised and semi-supervised clustering algorithms.
{"title":"Two Novel Kernel-Based Semi-Supervised Clustering Methods by Seeding","authors":"Lei Gu, F. Sun","doi":"10.1109/CCPR.2009.5344157","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344157","url":null,"abstract":"Semi-supervised clustering takes advantage of a small amount of labeled data to bring a great benefit to the clustering of unlabeled data. Based on a novel kernel method for clustering using one-class support vector machine, this paper presents two novel kernel-based semi-supervised clustering methods inspired by two semi-supervised variants of the k-means clustering algorithm by seeding respectively. To investigate the effectiveness of our approaches, experiments are done on three real datasets. Experimental results show that the proposed methods can improve the clustering performance significantly compared to other unsupervised and semi-supervised clustering algorithms.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"126 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117343325","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5343971
Seonghun Lee, Jae-Hyun Seok, Kyungmin Min, Jinhyung Kim
Robust extraction of text from scene images is essential for successful scene text recognition. Scene images usually have non- uniform illumination, complex background, and text-like objects. In this paper, we propose a text extraction algorithm by combining the adaptive binarization and perceptual color clustering method. Adaptive binarization method can handle gradual illumination changes on character regions, so it can extract whole character regions even though shadows and/or light variations affect the image quality. However, image binarization on gray-scale images cannot distinguish different color components having the same luminance. Perceptual color clustering method complementary can extract text regions which have similar color distances, so that it can prevent the problem of the binarization method. Text verification based on local information of a single component and global relationship between multiple components is used to determine the true text components. It is demonstrated that the proposed method achieved reasonabe accuracy of the text extraction for the moderately difficult examples from the ICDAR 2003 database.
{"title":"Scene Text Extraction Using Image Intensity and Color Information","authors":"Seonghun Lee, Jae-Hyun Seok, Kyungmin Min, Jinhyung Kim","doi":"10.1109/CCPR.2009.5343971","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5343971","url":null,"abstract":"Robust extraction of text from scene images is essential for successful scene text recognition. Scene images usually have non- uniform illumination, complex background, and text-like objects. In this paper, we propose a text extraction algorithm by combining the adaptive binarization and perceptual color clustering method. Adaptive binarization method can handle gradual illumination changes on character regions, so it can extract whole character regions even though shadows and/or light variations affect the image quality. However, image binarization on gray-scale images cannot distinguish different color components having the same luminance. Perceptual color clustering method complementary can extract text regions which have similar color distances, so that it can prevent the problem of the binarization method. Text verification based on local information of a single component and global relationship between multiple components is used to determine the true text components. It is demonstrated that the proposed method achieved reasonabe accuracy of the text extraction for the moderately difficult examples from the ICDAR 2003 database.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124192692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344022
Jinling Wang, Guiping Zhang, Na Ye, Lanhai Zhou
Term is the important component of the technical literature; automatic translation methods study on it has important research significance for international technology exchange. Under the background of the machine translation of the Japanese-Chinese patent documents and mainly study the term translation methods, this paper proposes a novel approach for Japanese-Chinese term translation based on multi-features on the basis of the existing IBM-model 4. Our approach effectively utilizes the features of the field attribute of the term and the character similarity between Japanese and Chinese, optimizes and improves the term translation results. The experimental results indicate that our method can make the precision of the term translation improvement 5.5%.
{"title":"Research on Japanese-Chinese Term Translation Technique Based on Multi-Features","authors":"Jinling Wang, Guiping Zhang, Na Ye, Lanhai Zhou","doi":"10.1109/CCPR.2009.5344022","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344022","url":null,"abstract":"Term is the important component of the technical literature; automatic translation methods study on it has important research significance for international technology exchange. Under the background of the machine translation of the Japanese-Chinese patent documents and mainly study the term translation methods, this paper proposes a novel approach for Japanese-Chinese term translation based on multi-features on the basis of the existing IBM-model 4. Our approach effectively utilizes the features of the field attribute of the term and the character similarity between Japanese and Chinese, optimizes and improves the term translation results. The experimental results indicate that our method can make the precision of the term translation improvement 5.5%.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"131 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124252972","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344122
Baojiang Zhong, K. Ma, Wenzhong Liu
Scale space techniques have attracted much attention in the field of computer vision and image processing. In particular, the curvature scale space (CSS) technique was selected in the MPEG-7 standard due to a number of nice properties. In this paper the scale space concept is first explained in detail and its significance in solving shape-based vision problems is clarified. A brief survey of the theoretical developments of the CSS technique in the past few decades is then presented.
{"title":"Curvature Scale Space Technique in Computer Vision: Basic Concept and Theoretical Developments","authors":"Baojiang Zhong, K. Ma, Wenzhong Liu","doi":"10.1109/CCPR.2009.5344122","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344122","url":null,"abstract":"Scale space techniques have attracted much attention in the field of computer vision and image processing. In particular, the curvature scale space (CSS) technique was selected in the MPEG-7 standard due to a number of nice properties. In this paper the scale space concept is first explained in detail and its significance in solving shape-based vision problems is clarified. A brief survey of the theoretical developments of the CSS technique in the past few decades is then presented.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"1 4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116781300","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：