Pub Date : 2012-07-15DOI: 10.1109/ICMLC.2012.6359584
Wei Wang, Jing Pan
Precise hand segmentation is crucial for gesture-based Human-Machine Interaction. Skin color based hand segmentation using skin color models shows poor performance in complex background where similar colors of the skin and non-uniform illumination exist. We propose a new method for hand segmentation by using an adaptive skin color model and the background information around the hand. Firstly, our method captures pixel values of the hand and the background then converts them into YCbCr color space. Secondly, skin and background Gaussian models based on the color space of CbCr are proposed. Lastly, these models are taken to segment the whole image respectively, and then required for the intersection. The main contribution of the paper is that the background information is taken into account to split image in reversed side to enhance the performance. Experimental results show that our method outperforms the method that uses the skin color model only.
{"title":"Hand segmentation using skin color and background information","authors":"Wei Wang, Jing Pan","doi":"10.1109/ICMLC.2012.6359584","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6359584","url":null,"abstract":"Precise hand segmentation is crucial for gesture-based Human-Machine Interaction. Skin color based hand segmentation using skin color models shows poor performance in complex background where similar colors of the skin and non-uniform illumination exist. We propose a new method for hand segmentation by using an adaptive skin color model and the background information around the hand. Firstly, our method captures pixel values of the hand and the background then converts them into YCbCr color space. Secondly, skin and background Gaussian models based on the color space of CbCr are proposed. Lastly, these models are taken to segment the whole image respectively, and then required for the intersection. The main contribution of the paper is that the background information is taken into account to split image in reversed side to enhance the performance. Experimental results show that our method outperforms the method that uses the skin color model only.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114142513","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 : 2012-07-15DOI: 10.1109/ICMLC.2012.6358921
Suyun Zhao, Si Lin
Currently, most works on interval valued problems mainly focus on attribute reduction (i.e., feature selection) by using rough set technologies. However, less research work on classifier building on interval-valued problems has been conducted. It is promising to propose an approach to build classifier for interval-valued problems. In this paper, we propose a classification approach based on interval valued fuzzy rough sets. First, the concept of interval valued fuzzy granules are proposed, which is the crucial notion to build the reduction framework for the interval-valued databases. Second, the idea to keep the critical value invariant before and after reduction is selected. Third, the structure of reduction rule is completely studied by using the discernibility vector approach. After the description of rule inference system, a set of rules covering all the objects can be obtained, which is used as a rule based classifier for future classification. Finally, numerical examples are presented to illustrate feasibility and affectivity of the proposed method in the application of privacy protection.
{"title":"Interval valued fuzzy rough classifier and its application on privacy protection","authors":"Suyun Zhao, Si Lin","doi":"10.1109/ICMLC.2012.6358921","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6358921","url":null,"abstract":"Currently, most works on interval valued problems mainly focus on attribute reduction (i.e., feature selection) by using rough set technologies. However, less research work on classifier building on interval-valued problems has been conducted. It is promising to propose an approach to build classifier for interval-valued problems. In this paper, we propose a classification approach based on interval valued fuzzy rough sets. First, the concept of interval valued fuzzy granules are proposed, which is the crucial notion to build the reduction framework for the interval-valued databases. Second, the idea to keep the critical value invariant before and after reduction is selected. Third, the structure of reduction rule is completely studied by using the discernibility vector approach. After the description of rule inference system, a set of rules covering all the objects can be obtained, which is used as a rule based classifier for future classification. Finally, numerical examples are presented to illustrate feasibility and affectivity of the proposed method in the application of privacy protection.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114231364","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 : 2012-07-15DOI: 10.1109/ICMLC.2012.6359479
Yi-Hua Ma, Dong-Li Zhang
Inspired by previous existing works, based on the local preferential redistribution rule of the load and the non-linear relation between load and capacity, we put forward a cascading model which is more practical and more suitable for real networks. We analyze the model theoretically and simulate it on BA scale-free network. In comparison with the strongest robustness against cascading failures of the linear load-capacity model in the case of α = 0.5 , we find that the robustness of the network can reach stronger in the case of δ ≠ 1, which is a tunable parameter controlling the strength of the capacity of node in our model. The results show that the model is effective. So it may be helpful to control cascading network failure and research on cascading failure deeply.
{"title":"Cascading network failure based on local load distribution and non-linear relationship between initial load and capacity","authors":"Yi-Hua Ma, Dong-Li Zhang","doi":"10.1109/ICMLC.2012.6359479","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6359479","url":null,"abstract":"Inspired by previous existing works, based on the local preferential redistribution rule of the load and the non-linear relation between load and capacity, we put forward a cascading model which is more practical and more suitable for real networks. We analyze the model theoretically and simulate it on BA scale-free network. In comparison with the strongest robustness against cascading failures of the linear load-capacity model in the case of α = 0.5 , we find that the robustness of the network can reach stronger in the case of δ ≠ 1, which is a tunable parameter controlling the strength of the capacity of node in our model. The results show that the model is effective. So it may be helpful to control cascading network failure and research on cascading failure deeply.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121158792","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 : 2012-07-15DOI: 10.1109/ICMLC.2012.6359596
Shing‐Tai Pan, Ching-Fa Chen, Ying-Wei Lee
A genetic algorithm is used to train the fuzzy membership function of a fuzzy codebook for the modeling of Discrete Hidden Markov Model (DHMM) applied to Mandarin speech recognition. Vector quantization for a speech feature based on a codebook is a fundamental process to recognize the speech signal by DHMM. A codebook with fuzzy membership functions corresponding to each vector in the codebook will be first trained by genetic algorithms (GAs) through speech features. The trained fuzzy codebook is then used to quantize the speech features. Subsequently, the quantized speech statistical features are used to model the DHMM for each speech. Besides, all the speech features to be recognized will go through the fuzzy codebook for quantization before being fed into the DHMM model for recognition. Experimental results show that both the speech recognition rate and computation time for recognition can be improved by the proposed strategy.
{"title":"Genetic algorithm on fuzzy codebook training for speech recognition","authors":"Shing‐Tai Pan, Ching-Fa Chen, Ying-Wei Lee","doi":"10.1109/ICMLC.2012.6359596","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6359596","url":null,"abstract":"A genetic algorithm is used to train the fuzzy membership function of a fuzzy codebook for the modeling of Discrete Hidden Markov Model (DHMM) applied to Mandarin speech recognition. Vector quantization for a speech feature based on a codebook is a fundamental process to recognize the speech signal by DHMM. A codebook with fuzzy membership functions corresponding to each vector in the codebook will be first trained by genetic algorithms (GAs) through speech features. The trained fuzzy codebook is then used to quantize the speech features. Subsequently, the quantized speech statistical features are used to model the DHMM for each speech. Besides, all the speech features to be recognized will go through the fuzzy codebook for quantization before being fed into the DHMM model for recognition. Experimental results show that both the speech recognition rate and computation time for recognition can be improved by the proposed strategy.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115782829","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 : 2012-07-15DOI: 10.1109/ICMLC.2012.6358889
Guoqing Chao, Shiliang Sun
This paper extracts seven effective feature sets and reduces them to same dimension by principle component analysis (peA), such that it can utilize a multitask feature sparsity approach to the automatic identification of semantic relations between nominals in English sentences under maximum entropy discrimination (MED) framework. This method can make full use of related information between different semantic classifications to perform multitask discriminative learning and don't employ additional knowledge sources. At SemEval 2007, our system achieved a F-score of 69.15 % which is higher than that by independent SVM.
{"title":"Applying a multitask feature sparsity method for the classification of semantic relations between nominals","authors":"Guoqing Chao, Shiliang Sun","doi":"10.1109/ICMLC.2012.6358889","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6358889","url":null,"abstract":"This paper extracts seven effective feature sets and reduces them to same dimension by principle component analysis (peA), such that it can utilize a multitask feature sparsity approach to the automatic identification of semantic relations between nominals in English sentences under maximum entropy discrimination (MED) framework. This method can make full use of related information between different semantic classifications to perform multitask discriminative learning and don't employ additional knowledge sources. At SemEval 2007, our system achieved a F-score of 69.15 % which is higher than that by independent SVM.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"174 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115955998","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 : 2012-07-15DOI: 10.1109/ICMLC.2012.6358972
Junjie Hu
In lots of important applications, such as malignant cell detection, network intrusion detection, error signal detection in power system, the data distributions of positive and negative classes are usually imbalance. Many classifiers could not perform well in data imbalance cases. The major problem is that classifiers tend to ignore samples and accuracy of the minority class without regarding the higher cost of misclassification in this minor class. Therefore, pattern classification for imbalance data becomes a hot challenge to both academy and industry. In this paper, we propose an active learning method for imbalance data using a stochastic sensitivity measure (ST-SM) of Radial Basis Function Neural Network (RBFNN). A large ST-SM indicates the RBFNN is uncertain and yields a large output fluctuation around a particular sample. These samples yielding large ST-SM values are selected for adding to the training set in each turn. Empirically, samples with large output perturbation (i.e. large ST-SM) should be located near the classification boundary and is of great significance for the training of classifier. As for the imbalance characteristic of the data set, the ST-SM should be able to reduce the number of redundant samples being selected in the majority class, rebalance the sample distribution of the training set, and finally improve the performance of the classifier.
{"title":"Active learning for imbalance problem using L-GEM of RBFNN","authors":"Junjie Hu","doi":"10.1109/ICMLC.2012.6358972","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6358972","url":null,"abstract":"In lots of important applications, such as malignant cell detection, network intrusion detection, error signal detection in power system, the data distributions of positive and negative classes are usually imbalance. Many classifiers could not perform well in data imbalance cases. The major problem is that classifiers tend to ignore samples and accuracy of the minority class without regarding the higher cost of misclassification in this minor class. Therefore, pattern classification for imbalance data becomes a hot challenge to both academy and industry. In this paper, we propose an active learning method for imbalance data using a stochastic sensitivity measure (ST-SM) of Radial Basis Function Neural Network (RBFNN). A large ST-SM indicates the RBFNN is uncertain and yields a large output fluctuation around a particular sample. These samples yielding large ST-SM values are selected for adding to the training set in each turn. Empirically, samples with large output perturbation (i.e. large ST-SM) should be located near the classification boundary and is of great significance for the training of classifier. As for the imbalance characteristic of the data set, the ST-SM should be able to reduce the number of redundant samples being selected in the majority class, rebalance the sample distribution of the training set, and finally improve the performance of the classifier.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116704245","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 : 2012-07-15DOI: 10.1109/ICMLC.2012.6359582
Yang Zhao, Wei Lu, Yan Zheng, Jian Wang
It has long been a big challenge to extract dense smoke regions by motion detection. As a result, there are too few suspected smoke regions being recognized for an early fire alarm. In this paper, an early smoke detecting system that can efficiently extract dense smoke regions is proposed. Firstly, since the brightness in the areas that have dynamic texture is not constant, the residuals of optical flow are calculated to locate suspected smoke regions. A certain threshold of the increment of optical flow residuals is also used to distinguish smoke from other dynamic texture. Secondly, five features that can jointly represent a smoke area, including grayish color, chrominance decrease, edge energy decrease, optical flow orientation diffusion and circularity, are chosen by thorough experiments. Experimental results show that the proposed system can detect the smoke in early time and is robust to most kinds of interferences, especially other dynamic textures.
{"title":"An early smoke detection system based on increment of optical flow residual","authors":"Yang Zhao, Wei Lu, Yan Zheng, Jian Wang","doi":"10.1109/ICMLC.2012.6359582","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6359582","url":null,"abstract":"It has long been a big challenge to extract dense smoke regions by motion detection. As a result, there are too few suspected smoke regions being recognized for an early fire alarm. In this paper, an early smoke detecting system that can efficiently extract dense smoke regions is proposed. Firstly, since the brightness in the areas that have dynamic texture is not constant, the residuals of optical flow are calculated to locate suspected smoke regions. A certain threshold of the increment of optical flow residuals is also used to distinguish smoke from other dynamic texture. Secondly, five features that can jointly represent a smoke area, including grayish color, chrominance decrease, edge energy decrease, optical flow orientation diffusion and circularity, are chosen by thorough experiments. Experimental results show that the proposed system can detect the smoke in early time and is robust to most kinds of interferences, especially other dynamic textures.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"61 6","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"113938186","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 : 2012-07-15DOI: 10.1109/ICMLC.2012.6359011
Xiaolian Guo, Haiying Wang, D. H. Glass
An extended Bayesian self-organizing map (BSOM) learning process is proposed, called the growing BSOM (GBSOM). It starts with two neurons and adds new neurons to the network via a process in which the neuron with the lowest individual log-likelihood is identified. It can automatically terminate and find the optimal number of neurons to represent the given dataset during the learning process. In this paper, three synthetic datasets and one real dataset are used to test the proposed algorithm, and three stopping criteria are used to automatically terminate the learning process, which are Bayesian information criterion (BIC) and two clustering validity indices: DB-Index and SV-Index. According to the results, using BIC as stopping criterion is better than using DB-Index and SV-Index as stopping criteria.
{"title":"A growing Bayesian self-organizing map for data clustering","authors":"Xiaolian Guo, Haiying Wang, D. H. Glass","doi":"10.1109/ICMLC.2012.6359011","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6359011","url":null,"abstract":"An extended Bayesian self-organizing map (BSOM) learning process is proposed, called the growing BSOM (GBSOM). It starts with two neurons and adds new neurons to the network via a process in which the neuron with the lowest individual log-likelihood is identified. It can automatically terminate and find the optimal number of neurons to represent the given dataset during the learning process. In this paper, three synthetic datasets and one real dataset are used to test the proposed algorithm, and three stopping criteria are used to automatically terminate the learning process, which are Bayesian information criterion (BIC) and two clustering validity indices: DB-Index and SV-Index. According to the results, using BIC as stopping criterion is better than using DB-Index and SV-Index as stopping criteria.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":" 12","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"113951383","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 : 2012-07-15DOI: 10.1109/ICMLC.2012.6359640
J. Lin, Jen-Yuan Yeh, Chao-Chung Liu
Information retrieval (IR) returns a relative ranking of documents with respect to a user query. Learning to rank for information retrieval (LR4IR) employs supervised learning techniques to address this problem, and it aims to produce a ranking model automatically for defining a proper sequential order of related documents based on the query. The ranking model determines the relationship degree between documents and the query. In this paper an improved version of RankGP is proposed. It uses layered multi-population genetic programming to obtain a ranking function which consists of a set of IR evidences and particular predefined operators. The proposed method is capable to generate complex functions through evolving small populations. In this paper, LETOR 4.0 was used to evaluate the effectiveness of the proposed method and the results showed that the method is competitive with other LR4IR Algorithms.
{"title":"Applying layered multi-population genetic programming on learning to rank for information retrieval","authors":"J. Lin, Jen-Yuan Yeh, Chao-Chung Liu","doi":"10.1109/ICMLC.2012.6359640","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6359640","url":null,"abstract":"Information retrieval (IR) returns a relative ranking of documents with respect to a user query. Learning to rank for information retrieval (LR4IR) employs supervised learning techniques to address this problem, and it aims to produce a ranking model automatically for defining a proper sequential order of related documents based on the query. The ranking model determines the relationship degree between documents and the query. In this paper an improved version of RankGP is proposed. It uses layered multi-population genetic programming to obtain a ranking function which consists of a set of IR evidences and particular predefined operators. The proposed method is capable to generate complex functions through evolving small populations. In this paper, LETOR 4.0 was used to evaluate the effectiveness of the proposed method and the results showed that the method is competitive with other LR4IR Algorithms.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"66 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124938760","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 : 2012-07-15DOI: 10.1109/ICMLC.2012.6359579
Shyi-Ming Chen, Wen-Chyuan Hsin, Yu-Chuan Chang
In this paper, we present a weighted fuzzy interpolative reasoning method based on the proposed PSO-based weights-learning algorithm. We also apply the proposed method to deal with the computer activity prediction problem. The experimental results show that the proposed weighted fuzzy interpolative reasoning method using the optimally learned weights obtained by the proposed PSO-based weights-learning algorithm gets smaller relative squared error rates than the existing methods.
{"title":"A new method for weighted fuzzy interpolative reasoning based on PSO-based weights-learning techniques","authors":"Shyi-Ming Chen, Wen-Chyuan Hsin, Yu-Chuan Chang","doi":"10.1109/ICMLC.2012.6359579","DOIUrl":"https://doi.org/10.1109/ICMLC.2012.6359579","url":null,"abstract":"In this paper, we present a weighted fuzzy interpolative reasoning method based on the proposed PSO-based weights-learning algorithm. We also apply the proposed method to deal with the computer activity prediction problem. The experimental results show that the proposed weighted fuzzy interpolative reasoning method using the optimally learned weights obtained by the proposed PSO-based weights-learning algorithm gets smaller relative squared error rates than the existing methods.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121481162","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：