Currently, optical device, such as microscopes and CCD cameras, are utilized for identification of tool marks in the field of forensic science which mainly depend on the experience of forensic scientists. A new approach using extended fractal analysis technology to classify tool marks such as striation patterns is presented. it computes four directional multi-scale extended fractal parameters and the maximum direction fractal feature, then performs a supervised classification. Experimental results demonstrate that this method provides a classification scheme that performs well than the traditional schemes and is effective for classification of tool marks.
{"title":"Striation Patterns Classification of Tool Marks Based on Extended Fractal Analysis","authors":"Min Yang, Donghong Li, Li Mou, Wei-dong Wang","doi":"10.1109/CCPR.2008.93","DOIUrl":"https://doi.org/10.1109/CCPR.2008.93","url":null,"abstract":"Currently, optical device, such as microscopes and CCD cameras, are utilized for identification of tool marks in the field of forensic science which mainly depend on the experience of forensic scientists. A new approach using extended fractal analysis technology to classify tool marks such as striation patterns is presented. it computes four directional multi-scale extended fractal parameters and the maximum direction fractal feature, then performs a supervised classification. Experimental results demonstrate that this method provides a classification scheme that performs well than the traditional schemes and is effective for classification of tool marks.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"105 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130542637","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}
With the increase of information on Internet, how to gain useful information fleetly and effectively becomes an important task, and information automatic classification emerges as the times require. Bayes has been used in many fields as one of the classification methods. This paper applies the classification model which Bayes classifier combines with language model to Chinese text classification. On the Chinese Corpus of FuDan University, our experiments show that the improved classifiers which used the four smoothing methods have better performance than naive Bayes classifier model. In particular with the method Jelinek-Mercer of adopting modified smoothing scale, the performance of classifier improves a lot.
{"title":"Research on the Methods of Chinese Text Classification using Bayes and Language Model","authors":"Tao Yan, Guangyong Gao","doi":"10.1109/CCPR.2008.88","DOIUrl":"https://doi.org/10.1109/CCPR.2008.88","url":null,"abstract":"With the increase of information on Internet, how to gain useful information fleetly and effectively becomes an important task, and information automatic classification emerges as the times require. Bayes has been used in many fields as one of the classification methods. This paper applies the classification model which Bayes classifier combines with language model to Chinese text classification. On the Chinese Corpus of FuDan University, our experiments show that the improved classifiers which used the four smoothing methods have better performance than naive Bayes classifier model. In particular with the method Jelinek-Mercer of adopting modified smoothing scale, the performance of classifier improves a lot.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129847147","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}
A novel color correlogram based particle filter was proposed for an object tracking in visual surveillance. By using the color correlogram as object feature, spatial information is incorporated into object representation, which yields a reliable likelihood description of the observation and prediction for tracking the objects accurately. The capability of the tracker to tolerate appearance changes like orientation changes, small scale changes, partial occlusions and background scene changes is demonstrated using real image sequences. Experimental evidence shows that the color correlogram is more effective than the traditional color histogram for objects tracking.
{"title":"Color Correlogram Based Particle Filter for Object Tracking","authors":"Zhang Tao, S. Fei, Hong Lu, Xiaodong Li","doi":"10.1109/CCPR.2008.45","DOIUrl":"https://doi.org/10.1109/CCPR.2008.45","url":null,"abstract":"A novel color correlogram based particle filter was proposed for an object tracking in visual surveillance. By using the color correlogram as object feature, spatial information is incorporated into object representation, which yields a reliable likelihood description of the observation and prediction for tracking the objects accurately. The capability of the tracker to tolerate appearance changes like orientation changes, small scale changes, partial occlusions and background scene changes is demonstrated using real image sequences. Experimental evidence shows that the color correlogram is more effective than the traditional color histogram for objects tracking.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115321658","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}
In the age of Internet, with the online information explosive growth, people want to find information we need in the cyberworld fleetly and exactly. The information retrieval method based on the keyword or the simple logic-combination of the keywords has been unable to meet the people's need of information getting to a certain extent. Gradually intelligent question answering system has grown to satisfy the people's need. This paper revolves around design and implementation of intelligent question answering system in a restricted domain, does a series of research aiming at the construction of domain knowledge, questions' comprehension and analysis, FAQ question matching, and so on. The FAQ question match is implemented by sentence similarity computation, and this model can answer frequently-asked question fast and concisely. Besides the system constructs theme document library taking advantage of web pages which Web crawler fetches. For the question which can not be answered by FAQ, the system will find answers from the theme document library. That is supplement and perfection of question answering system.
{"title":"Research and Implementation of Intelligent Question Answering System in a Restricted Domain","authors":"Yinli Wang, Guanglai Gao","doi":"10.1109/CCPR.2008.89","DOIUrl":"https://doi.org/10.1109/CCPR.2008.89","url":null,"abstract":"In the age of Internet, with the online information explosive growth, people want to find information we need in the cyberworld fleetly and exactly. The information retrieval method based on the keyword or the simple logic-combination of the keywords has been unable to meet the people's need of information getting to a certain extent. Gradually intelligent question answering system has grown to satisfy the people's need. This paper revolves around design and implementation of intelligent question answering system in a restricted domain, does a series of research aiming at the construction of domain knowledge, questions' comprehension and analysis, FAQ question matching, and so on. The FAQ question match is implemented by sentence similarity computation, and this model can answer frequently-asked question fast and concisely. Besides the system constructs theme document library taking advantage of web pages which Web crawler fetches. For the question which can not be answered by FAQ, the system will find answers from the theme document library. That is supplement and perfection of question answering system.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131071004","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}
Gait recognition, a new biometrics recognition technology, can discriminate individuals by the way they walk. A novel gait recognition method based on standard deviation energy image is proposed in this paper. Firstly, it divides video sequences into several gait cycles. Secondly, two kinds of energy images called nonzero and zero standard deviation energy image respectively are constructed. Finally, features vectors of two energy images are extracted and fused in subspace by means of principal components analysis. Experimental results show that our approach is valid and has encouraging recognition performance.
{"title":"A Novel Gait Recognition Method Based on Standard Deviation Energy Image","authors":"Lei Gu, F. Sun","doi":"10.1109/CCPR.2008.64","DOIUrl":"https://doi.org/10.1109/CCPR.2008.64","url":null,"abstract":"Gait recognition, a new biometrics recognition technology, can discriminate individuals by the way they walk. A novel gait recognition method based on standard deviation energy image is proposed in this paper. Firstly, it divides video sequences into several gait cycles. Secondly, two kinds of energy images called nonzero and zero standard deviation energy image respectively are constructed. Finally, features vectors of two energy images are extracted and fused in subspace by means of principal components analysis. Experimental results show that our approach is valid and has encouraging recognition performance.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114019414","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}
In the large vocabulary continuous speech recognition system based on stochastic segment model (SSM), the multistage decoding and pruning algorithm could decrease decoding time obviously. Generally, we only decode and prune for one segment each time. In this paper, a decoding algorithm based on neighboring segments is proposed. This algorithm decodes for multi-segments at the same time, so that the threshold of every segment could be highly shared by all the segments in each stage. That means more useless computation would be avoided, and the decoding would become faster. When using this algorithm in LVCSR system, we saved the decoding time of 50% approximately without accuracy loss.
{"title":"Stochastic Segment Model Decoding Algorithm Based on Neighboring Segments and its Application in LVCSR","authors":"Shouye Peng, Wenju Liu, Huayun Zhang","doi":"10.1109/CCPR.2008.90","DOIUrl":"https://doi.org/10.1109/CCPR.2008.90","url":null,"abstract":"In the large vocabulary continuous speech recognition system based on stochastic segment model (SSM), the multistage decoding and pruning algorithm could decrease decoding time obviously. Generally, we only decode and prune for one segment each time. In this paper, a decoding algorithm based on neighboring segments is proposed. This algorithm decodes for multi-segments at the same time, so that the threshold of every segment could be highly shared by all the segments in each stage. That means more useless computation would be avoided, and the decoding would become faster. When using this algorithm in LVCSR system, we saved the decoding time of 50% approximately without accuracy loss.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114952378","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}
As a supervised learning algorithm, the standard Gaussian processes has the excellent performance of classification. In this paper, we present a semi-supervised algorithm to learning a Gaussian process classifier, which incorporating a graph-based construction of semi-supervised kernels in the presence of labeled and unlabeled data, and expanding the standard Gaussian processes algorithm into the semi-supervised learning framework. Our algorithm adopts the spectral decomposition to obtain the kernel matrices, and employs a convex optimization method to learn an optimal semi-supervised kernel, which is incorporated into the Gaussian process model. In the Gaussian processes classification, the expectation propagation algorithm is applied to approximate the Gaussian posterior distribution. The main characteristic of the proposed algorithm is that we incorporate the geometric properties of unlabeled data by globally defined kernel functions. The semi-supervised Gaussian processes model has an explicitly probabilistic interpretation, and can model the uncertainty among the data and solve the complex non-linear inference problems. In the presence of few labeled examples, the proposed algorithm outperforms cross-validation methods, and we present the experimental results demonstrating the effectiveness of this algorithm in comparison with other related works in the literature.
{"title":"Semi-Supervised Learning with Gaussian Processes","authors":"Hongwei Li, Yakui Li, Hanqing Lu","doi":"10.1109/CCPR.2008.12","DOIUrl":"https://doi.org/10.1109/CCPR.2008.12","url":null,"abstract":"As a supervised learning algorithm, the standard Gaussian processes has the excellent performance of classification. In this paper, we present a semi-supervised algorithm to learning a Gaussian process classifier, which incorporating a graph-based construction of semi-supervised kernels in the presence of labeled and unlabeled data, and expanding the standard Gaussian processes algorithm into the semi-supervised learning framework. Our algorithm adopts the spectral decomposition to obtain the kernel matrices, and employs a convex optimization method to learn an optimal semi-supervised kernel, which is incorporated into the Gaussian process model. In the Gaussian processes classification, the expectation propagation algorithm is applied to approximate the Gaussian posterior distribution. The main characteristic of the proposed algorithm is that we incorporate the geometric properties of unlabeled data by globally defined kernel functions. The semi-supervised Gaussian processes model has an explicitly probabilistic interpretation, and can model the uncertainty among the data and solve the complex non-linear inference problems. In the presence of few labeled examples, the proposed algorithm outperforms cross-validation methods, and we present the experimental results demonstrating the effectiveness of this algorithm in comparison with other related works in the literature.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129914287","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}
Dealing with nonlinear deformation is one of the challenges in fingerprint matching. Nonlinear deformation, arising from the plasticity of the skin, and the pressure and the other factors during acquisition, lead to difficulties in establishing a match between multiple impressions from the same fingerprint. One solution to deal with the nonlinear deformation is to estimate a model to describe it, like thin-plate spline model (TPS). The match method based on model could be described as the following steps. First, local match algorithm is performed on input image and template image to obtain some corresponding minutiae. Then, the corresponding points are used to calculate the parameters of the model. Based on the model, two images are aligned and more minutiae was matched. Repeat the second and third steps until no more minutiae are matched. In the original TPS model, the differences between every two corresponding points from the local match are ignored. Meanwhile, minutiae orientation is not taken into account. In this paper a weighted thin-plate spline model using both the difference between corresponding points and the minutiae orientation was proposed to describe the fingerprint deformation, and the performance of the fingerprint match algorithm based on the new model on databases of FVC2004 turned to be efficient and robust.
{"title":"Weight Thin-Plate Spline Fingerprint Matching Using Minutiae Locations and Orientations","authors":"Dong-jin Fan, Zi-Rui Deng, Ju-fu Feng","doi":"10.1109/CCPR.2008.66","DOIUrl":"https://doi.org/10.1109/CCPR.2008.66","url":null,"abstract":"Dealing with nonlinear deformation is one of the challenges in fingerprint matching. Nonlinear deformation, arising from the plasticity of the skin, and the pressure and the other factors during acquisition, lead to difficulties in establishing a match between multiple impressions from the same fingerprint. One solution to deal with the nonlinear deformation is to estimate a model to describe it, like thin-plate spline model (TPS). The match method based on model could be described as the following steps. First, local match algorithm is performed on input image and template image to obtain some corresponding minutiae. Then, the corresponding points are used to calculate the parameters of the model. Based on the model, two images are aligned and more minutiae was matched. Repeat the second and third steps until no more minutiae are matched. In the original TPS model, the differences between every two corresponding points from the local match are ignored. Meanwhile, minutiae orientation is not taken into account. In this paper a weighted thin-plate spline model using both the difference between corresponding points and the minutiae orientation was proposed to describe the fingerprint deformation, and the performance of the fingerprint match algorithm based on the new model on databases of FVC2004 turned to be efficient and robust.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124570333","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}
The basic sampling importance resampling algorithm is the basic for improving particle filter methods which are widely utilized in optimal filtering problems. In our paper, we introduce a modified basic SIR algorithm and analyze the convergence property of the modified basic SIR algorithm. Furthermore, when the recursive time is finite and the forth-order moment of the interesting function w.r.t the posterior joint distribution of the extended state is exist, the sufficient condition for the basic particle filter estimation convergence almost surely to the optimal estimation is discussed.
{"title":"Convergence Properties of Particle Filter Algorithm","authors":"Yanwen Qu, Yi Chen, Jing-yu Yang","doi":"10.1109/CCPR.2008.14","DOIUrl":"https://doi.org/10.1109/CCPR.2008.14","url":null,"abstract":"The basic sampling importance resampling algorithm is the basic for improving particle filter methods which are widely utilized in optimal filtering problems. In our paper, we introduce a modified basic SIR algorithm and analyze the convergence property of the modified basic SIR algorithm. Furthermore, when the recursive time is finite and the forth-order moment of the interesting function w.r.t the posterior joint distribution of the extended state is exist, the sufficient condition for the basic particle filter estimation convergence almost surely to the optimal estimation is discussed.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128277073","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}
Zhonghua Liu, Zhong Jin, Zhihui Lai, Chuanbo Huang, M. Wan
Combined with wavelet transform, singular value decomposition and kernel principal component analysis, a method for face recognition is presented. Firstly, the wavelet transformation is used to reduce the dimension of the face picture. Then, SVD is used to subtract the features of the lowest resolution subimage, and the singular value feature vector is mapped onto the feature space with kpca and obtains nonlinear feature . Finally, face recognition can be realized according to BP neural network method. Experimental results on ORL and YALE face-databases show that the recognition rate by the proposed method is higher than that by KPCA, SVD, WT-KPCA and WT-SVD respectively.
{"title":"Face Recognition Based on Wavelet Transform, Singular Value Decomposition and Kernel Principal Component Analysis","authors":"Zhonghua Liu, Zhong Jin, Zhihui Lai, Chuanbo Huang, M. Wan","doi":"10.1109/CCPR.2008.61","DOIUrl":"https://doi.org/10.1109/CCPR.2008.61","url":null,"abstract":"Combined with wavelet transform, singular value decomposition and kernel principal component analysis, a method for face recognition is presented. Firstly, the wavelet transformation is used to reduce the dimension of the face picture. Then, SVD is used to subtract the features of the lowest resolution subimage, and the singular value feature vector is mapped onto the feature space with kpca and obtains nonlinear feature . Finally, face recognition can be realized according to BP neural network method. Experimental results on ORL and YALE face-databases show that the recognition rate by the proposed method is higher than that by KPCA, SVD, WT-KPCA and WT-SVD respectively.","PeriodicalId":292956,"journal":{"name":"2008 Chinese Conference on Pattern Recognition","volume":"73 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130840162","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：