Faced with the evolving attacks in recommender systems, many detection features have been proposed by human engineering and used in supervised or unsupervised detection methods. However, the detection features extract...Faced with the evolving attacks in recommender systems, many detection features have been proposed by human engineering and used in supervised or unsupervised detection methods. However, the detection features extracted by human engineering are usually aimed at some specific types of attacks. To further detect other new types of attacks, the traditional methods have to re-extract detection features with high knowledge cost. To address these limitations, the method for automatic extraction of robust features is proposed and then an Adaboost-based detection method is presented. Firstly, to obtain robust representation with prior knowledge, unlike uniform corruption rate in traditional mLDA(marginalized Linear Denoising Autoencoder), different corruption rates for items are calculated according to the ratings’ distribution. Secondly, the ratings sparsity is used to weight the mapping matrix to extract low-dimensional representation. Moreover, the uniform corruption rate is also set to the next layer in mSLDA(marginalized Stacked Linear Denoising Autoencoder) to extract the stable and robust user features. Finally, under the robust feature space, an Adaboost-based detection method is proposed to alleviate the imbalanced classification problem. Experimental results on the Netflix and Amazon review datasets indicate that the proposed method can effectively detect various attacks.展开更多
Wake-Up-Word Speech Recognition task (WUW-SR) is a computationally very demand, particularly the stage of feature extraction which is decoded with corresponding Hidden Markov Models (HMMs) in the back-end stage of the...Wake-Up-Word Speech Recognition task (WUW-SR) is a computationally very demand, particularly the stage of feature extraction which is decoded with corresponding Hidden Markov Models (HMMs) in the back-end stage of the WUW-SR. The state of the art WUW-SR system is based on three different sets of features: Mel-Frequency Cepstral Coefficients (MFCC), Linear Predictive Coding Coefficients (LPC), and Enhanced Mel-Frequency Cepstral Coefficients (ENH_MFCC). In (front-end of Wake-Up-Word Speech Recognition System Design on FPGA) [1], we presented an experimental FPGA design and implementation of a novel architecture of a real-time spectrogram extraction processor that generates MFCC, LPC, and ENH_MFCC spectrograms simultaneously. In this paper, the details of converting the three sets of spectrograms 1) Mel-Frequency Cepstral Coefficients (MFCC), 2) Linear Predictive Coding Coefficients (LPC), and 3) Enhanced Mel-Frequency Cepstral Coefficients (ENH_MFCC) to their equivalent features are presented. In the WUW- SR system, the recognizer’s frontend is located at the terminal which is typically connected over a data network to remote back-end recognition (e.g., server). The WUW-SR is shown in Figure 1. The three sets of speech features are extracted at the front-end. These extracted features are then compressed and transmitted to the server via a dedicated channel, where subsequently they are decoded.展开更多
The extraction of water bodies is essential for monitoring water resources,ecosystem services and the hydrological cycle,so analyzing water bodies from remote sensing images is necessary.The water index is designed to...The extraction of water bodies is essential for monitoring water resources,ecosystem services and the hydrological cycle,so analyzing water bodies from remote sensing images is necessary.The water index is designed to highlight water bodies in remote sensing images.We employ a new water index and digital image processing technology to extract water bodies automatically and accurately from Landsat 8 OLI images.Firstly,we preprocess Landsat 8 OLI images with radiometric calibration and atmospheric correction.Subsequently,we apply KT transformation,LBV transformation,AWEI nsh,and HIS transformation to the preprocessed image to calculate a new water index.Then,we perform linear feature enhancement and improve the local adaptive threshold segmentation method to extract small water bodies accurately.Meanwhile,we employ morphological enhancement and improve the local adaptive threshold segmentation method to extract large water bodies.Finally,we combine small and large water bodies to get complete water bodies.Compared with other traditional methods,our method has apparent advantages in water extraction,particularly in the extraction of small water bodies.展开更多
In recent years, the accuracy of speech recognition (SR) has been one of the most active areas of research. Despite that SR systems are working reasonably well in quiet conditions, they still suffer severe performance...In recent years, the accuracy of speech recognition (SR) has been one of the most active areas of research. Despite that SR systems are working reasonably well in quiet conditions, they still suffer severe performance degradation in noisy conditions or distorted channels. It is necessary to search for more robust feature extraction methods to gain better performance in adverse conditions. This paper investigates the performance of conventional and new hybrid speech feature extraction algorithms of Mel Frequency Cepstrum Coefficient (MFCC), Linear Prediction Coding Coefficient (LPCC), perceptual linear production (PLP), and RASTA-PLP in noisy conditions through using multivariate Hidden Markov Model (HMM) classifier. The behavior of the proposal system is evaluated using TIDIGIT human voice dataset corpora, recorded from 208 different adult speakers in both training and testing process. The theoretical basis for speech processing and classifier procedures were presented, and the recognition results were obtained based on word recognition rate.展开更多
A direct linear discriminant analysis algorithm based on economic singular value decomposition （DLDA/ESVD） is proposed to address the computationally complex problem of the conventional DLDA algorithm, which directl...A direct linear discriminant analysis algorithm based on economic singular value decomposition （DLDA/ESVD） is proposed to address the computationally complex problem of the conventional DLDA algorithm, which directly uses ESVD to reduce dimension and extract eigenvectors corresponding to nonzero eigenvalues. Then a DLDA algorithm based on column pivoting orthogonal triangular （QR） decomposition and ESVD （DLDA/QR-ESVD） is proposed to improve the performance of the DLDA/ESVD algorithm by processing a high-dimensional low rank matrix, which uses column pivoting QR decomposition to reduce dimension and ESVD to extract eigenvectors corresponding to nonzero eigenvalues. The experimental results on ORL, FERET and YALE face databases show that the proposed two algorithms can achieve almost the same performance and outperform the conventional DLDA algorithm in terms of computational complexity and training time. In addition, the experimental results on random data matrices show that the DLDA/QR-ESVD algorithm achieves better performance than the DLDA/ESVD algorithm by processing high-dimensional low rank matrices.展开更多
Principal Component Analysis （PCA） and Linear Discriminant Analysis （LDA） are two popular feature extraction techniques in statistical pattern recognition field. Due to small sample size problem LDA cannot be dire...Principal Component Analysis （PCA） and Linear Discriminant Analysis （LDA） are two popular feature extraction techniques in statistical pattern recognition field. Due to small sample size problem LDA cannot be directly applied to appearance-based face recognition tasks. As a consequence, a lot of LDA-based facial feature extraction techniques are proposed to deal with the problem one after the other. Nullspace Method is one of the most effective methods among them. The Nullspace Method tries to find a set of discriminant vectors which maximize the between-class scatter in the null space of the within-class scatter matrix. The calculation of its discriminant vectors will involve performing singular value decomposition on a high-dimensional matrix. It is generally memory- and time-consuming. Borrowing the key idea in Nullspace method and the concept of coefficient of variance in statistical analysis we present a novel facial feature extraction method, i.e., Discriminant based on Coefficient of Variance （DCV） in this paper. Experimental results performed on the FERET and AR face image databases demonstrate that DCV is a promising technique in comparison with Eigenfaces, Nullspace Method, and other state-of-the-art facial feature extraction methods.展开更多
Visual process monitoring is important in complex chemical processes.To address the high state separation of industrial data,we propose a new criterion for feature extraction called balanced multiple weighted linear d...Visual process monitoring is important in complex chemical processes.To address the high state separation of industrial data,we propose a new criterion for feature extraction called balanced multiple weighted linear discriminant analysis(BMWLDA).Then,we combine BMWLDA with self-organizing map(SOM)for visual monitoring of industrial operation processes.BMWLDA can extract the discriminative feature vectors from the original industrial data and maximally separate industrial operation states in the space spanned by these discriminative feature vectors.When the discriminative feature vectors are used as the input to SOM,the training result of SOM can differentiate industrial operation states clearly.This function improves the performance of visual monitoring.Continuous stirred tank reactor is used to verify that the class separation performance of BMWLDA is more effective than that of traditional linear discriminant analysis,approximate pairwise accuracy criterion,max–min distance analysis,maximum margin criterion,and local Fisher discriminant analysis.In addition,the method that combines BMWLDA with SOM can effectively perform visual process monitoring in real time.展开更多
Capturing the distributed platform with remotely controlled compromised machines using botnet is extensively analyzed by various researchers.However,certain limitations need to be addressed efficiently.The provisionin...Capturing the distributed platform with remotely controlled compromised machines using botnet is extensively analyzed by various researchers.However,certain limitations need to be addressed efficiently.The provisioning of detection mechanism with learning approaches provides a better solution more broadly by saluting multi-objective constraints.The bots’patterns or features over the network have to be analyzed in both linear and non-linear manner.The linear and non-linear features are composed of high-level and low-level features.The collected features are maintained over the Bag of Features(BoF)where the most influencing features are collected and provided into the classifier model.Here,the linearity and non-linearity of the threat are evaluated with Support Vector Machine(SVM).Next,with the collected BoF,the redundant features are eliminated as it triggers overhead towards the predictor model.Finally,a novel Incoming data Redundancy Elimination-based learning model(RedE-L)is built to classify the network features to provide robustness towards BotNets detection.The simulation is carried out in MATLAB environment,and the evaluation of proposed RedE-L model is performed with various online accessible network traffic dataset(benchmark dataset).The proposed model intends to show better tradeoff compared to the existing approaches like conventional SVM,C4.5,RepTree and so on.Here,various metrics like Accuracy,detection rate,Mathews Correlation Coefficient(MCC),and some other statistical analysis are performed to show the proposed RedE-L model's reliability.The F1-measure is 99.98%,precision is 99.93%,Accuracy is 99.84%,TPR is 99.92%,TNR is 99.94%,FNR is 0.06 and FPR is 0.06 respectively.展开更多
For the case where all multivariate normal parameters are known, we derive a new linear dimension reduction (LDR) method to determine a low-dimensional subspace that preserves or nearly preserves the original feature-...For the case where all multivariate normal parameters are known, we derive a new linear dimension reduction (LDR) method to determine a low-dimensional subspace that preserves or nearly preserves the original feature-space separation of the individual populations and the Bayes probability of misclassification. We also give necessary and sufficient conditions which provide the smallest reduced dimension that essentially retains the Bayes probability of misclassification from the original full-dimensional space in the reduced space. Moreover, our new LDR procedure requires no computationally expensive optimization procedure. Finally, for the case where parameters are unknown, we devise a LDR method based on our new theorem and compare our LDR method with three competing LDR methods using Monte Carlo simulations and a parametric bootstrap based on real data.展开更多
The recognition of pathological voice is considered a difficult task for speech analysis.Moreover,otolaryngologists needed to rely on oral communication with patients to discover traces of voice pathologies like dysph...The recognition of pathological voice is considered a difficult task for speech analysis.Moreover,otolaryngologists needed to rely on oral communication with patients to discover traces of voice pathologies like dysphonia that are caused by voice alteration of vocal folds and their accuracy is between 60%–70%.To enhance detection accuracy and reduce processing speed of dysphonia detection,a novel approach is proposed in this paper.We have leveraged Linear Discriminant Analysis(LDA)to train multiple Machine Learning(ML)models for dysphonia detection.Several ML models are utilized like Support Vector Machine(SVM),Logistic Regression,and K-nearest neighbor(K-NN)to predict the voice pathologies based on features like Mel-Frequency Cepstral Coefficients(MFCC),Fundamental Frequency(F0),Shimmer(%),Jitter(%),and Harmonic to Noise Ratio(HNR).The experiments were performed using Saarbrucken Voice Data-base(SVD)and a privately collected dataset.The K-fold cross-validation approach was incorporated to increase the robustness and stability of the ML models.According to the experimental results,our proposed approach has a 70%increase in processing speed over Principal Component Analysis(PCA)and performs remarkably well with a recognition accuracy of 95.24%on the SVD dataset surpassing the previous best accuracy of 82.37%.In the case of the private dataset,our proposed method achieved an accuracy rate of 93.37%.It can be an effective non-invasive method to detect dysphonia.展开更多
In order to achieve failure prediction without manual intervention for distributed systems, a novel failure feature analysis and extraction approach to automate failure prediction is proposed. Compared with the tradit...In order to achieve failure prediction without manual intervention for distributed systems, a novel failure feature analysis and extraction approach to automate failure prediction is proposed. Compared with the traditional methods which focus on building heuristic rules or models, the autonomic prediction approach analyzes the nonlinear correlation of failure features by recognizing failure patterns. Failure data are sorted according to the nonlinear correlation and failure signature is proposed for autonomic prediction. In addition, the Manifold Learning algorithm named supervised locally linear embedding is applied to achieve feature extraction. Based on the runtime monitoring of failure metrics, the experimental results indicate that the proposed method has better performance in terms of both correlation recognition precision and feature extraction quality and thus it can be used to design efficient autonomic failure prediction for distributed systems.展开更多
This paper introduces an idea of generating a kernel from an arbitrary function by embedding the training samples into the function.Based on this idea,we present two nonlinear feature extraction methods:generating ker...This paper introduces an idea of generating a kernel from an arbitrary function by embedding the training samples into the function.Based on this idea,we present two nonlinear feature extraction methods:generating kernel principal component analysis(GKPCA)and generating kernel Fisher discriminant(GKFD).These two methods are shown to be equivalent to the function-mapping-space PCA(FMS-PCA)and the function-mapping-space linear discriminant analysis(FMS-LDA)methods,respectively.This equivalence reveals that the generating kernel is actually determined by the corresponding function map.From the generating kernel point of view,we can classify the current kernel Fisher discriminant(KFD)algorithms into two categories:KPCA+LDA based algorithms and straightforward KFD(SKFD)algorithms.The KPCA+LDA based algorithms directly work on the given kernel and are not suitable for non-kernel functions,while the SKFD algorithms essentially work on the generating kernel from a given symmetric function and are therefore suitable for non-kernels as well as kernels.Finally,we outline the tensor-based feature extraction methods and discuss ways of extending tensor-based methods to their generating kernel versions.展开更多
局部线性嵌入算法采用欧氏距离选择邻域点,这通常会损失数据集本身的非线性特征,造成邻域点选取错误,且仅使用欧氏距离构造权重会导致信息挖掘不充分。针对以上问题,提出基于概率模型与信息熵的局部线性嵌入算法(Probability informatio...局部线性嵌入算法采用欧氏距离选择邻域点,这通常会损失数据集本身的非线性特征,造成邻域点选取错误,且仅使用欧氏距离构造权重会导致信息挖掘不充分。针对以上问题,提出基于概率模型与信息熵的局部线性嵌入算法(Probability information entropy-LLE,PIE-LLE)。首先,为了使邻域点选择更加合理,从数据集的概率分布角度出发,考虑样本点及其邻域的概率分布,为样本点构造符合局部分布的邻域集合。其次,为了充分提取样本的局部结构信息,在权重构造阶段,分别计算样本所属邻域概率以及每个样本的信息熵,融合二者信息重构低维样本。最后,在两个轴承故障数据集上的实验表明,所提方法故障识别准确度最高达到了100%,高于其他对比算法;在邻域点个数5~15范围内,PIE-LLE算法展现出良好的低维可视化效果;在参数敏感性实验中,该算法可以保持Fisher指标较大,有效提高了算法的分类准确度和稳定性。展开更多
基金supported by the National Natural Science Foundation of China [Nos. 61772452, 61379116]the Scientific and Technological Innovation Programs of Higher Education Institutions in Shanxi [No.2019L0847]the Natural Science Foundation of Hebei Province, China [No. F2015203046]
文摘Faced with the evolving attacks in recommender systems, many detection features have been proposed by human engineering and used in supervised or unsupervised detection methods. However, the detection features extracted by human engineering are usually aimed at some specific types of attacks. To further detect other new types of attacks, the traditional methods have to re-extract detection features with high knowledge cost. To address these limitations, the method for automatic extraction of robust features is proposed and then an Adaboost-based detection method is presented. Firstly, to obtain robust representation with prior knowledge, unlike uniform corruption rate in traditional mLDA(marginalized Linear Denoising Autoencoder), different corruption rates for items are calculated according to the ratings’ distribution. Secondly, the ratings sparsity is used to weight the mapping matrix to extract low-dimensional representation. Moreover, the uniform corruption rate is also set to the next layer in mSLDA(marginalized Stacked Linear Denoising Autoencoder) to extract the stable and robust user features. Finally, under the robust feature space, an Adaboost-based detection method is proposed to alleviate the imbalanced classification problem. Experimental results on the Netflix and Amazon review datasets indicate that the proposed method can effectively detect various attacks.
文摘Wake-Up-Word Speech Recognition task (WUW-SR) is a computationally very demand, particularly the stage of feature extraction which is decoded with corresponding Hidden Markov Models (HMMs) in the back-end stage of the WUW-SR. The state of the art WUW-SR system is based on three different sets of features: Mel-Frequency Cepstral Coefficients (MFCC), Linear Predictive Coding Coefficients (LPC), and Enhanced Mel-Frequency Cepstral Coefficients (ENH_MFCC). In (front-end of Wake-Up-Word Speech Recognition System Design on FPGA) [1], we presented an experimental FPGA design and implementation of a novel architecture of a real-time spectrogram extraction processor that generates MFCC, LPC, and ENH_MFCC spectrograms simultaneously. In this paper, the details of converting the three sets of spectrograms 1) Mel-Frequency Cepstral Coefficients (MFCC), 2) Linear Predictive Coding Coefficients (LPC), and 3) Enhanced Mel-Frequency Cepstral Coefficients (ENH_MFCC) to their equivalent features are presented. In the WUW- SR system, the recognizer’s frontend is located at the terminal which is typically connected over a data network to remote back-end recognition (e.g., server). The WUW-SR is shown in Figure 1. The three sets of speech features are extracted at the front-end. These extracted features are then compressed and transmitted to the server via a dedicated channel, where subsequently they are decoded.
基金Auhui Provincial Key Research and Development Project(No.202004a07020050)National Natural Science Foundation of China Youth Program(No.61901006)。
文摘The extraction of water bodies is essential for monitoring water resources,ecosystem services and the hydrological cycle,so analyzing water bodies from remote sensing images is necessary.The water index is designed to highlight water bodies in remote sensing images.We employ a new water index and digital image processing technology to extract water bodies automatically and accurately from Landsat 8 OLI images.Firstly,we preprocess Landsat 8 OLI images with radiometric calibration and atmospheric correction.Subsequently,we apply KT transformation,LBV transformation,AWEI nsh,and HIS transformation to the preprocessed image to calculate a new water index.Then,we perform linear feature enhancement and improve the local adaptive threshold segmentation method to extract small water bodies accurately.Meanwhile,we employ morphological enhancement and improve the local adaptive threshold segmentation method to extract large water bodies.Finally,we combine small and large water bodies to get complete water bodies.Compared with other traditional methods,our method has apparent advantages in water extraction,particularly in the extraction of small water bodies.
文摘In recent years, the accuracy of speech recognition (SR) has been one of the most active areas of research. Despite that SR systems are working reasonably well in quiet conditions, they still suffer severe performance degradation in noisy conditions or distorted channels. It is necessary to search for more robust feature extraction methods to gain better performance in adverse conditions. This paper investigates the performance of conventional and new hybrid speech feature extraction algorithms of Mel Frequency Cepstrum Coefficient (MFCC), Linear Prediction Coding Coefficient (LPCC), perceptual linear production (PLP), and RASTA-PLP in noisy conditions through using multivariate Hidden Markov Model (HMM) classifier. The behavior of the proposal system is evaluated using TIDIGIT human voice dataset corpora, recorded from 208 different adult speakers in both training and testing process. The theoretical basis for speech processing and classifier procedures were presented, and the recognition results were obtained based on word recognition rate.
基金The National Natural Science Foundation of China (No.61374194)
文摘A direct linear discriminant analysis algorithm based on economic singular value decomposition （DLDA/ESVD） is proposed to address the computationally complex problem of the conventional DLDA algorithm, which directly uses ESVD to reduce dimension and extract eigenvectors corresponding to nonzero eigenvalues. Then a DLDA algorithm based on column pivoting orthogonal triangular （QR） decomposition and ESVD （DLDA/QR-ESVD） is proposed to improve the performance of the DLDA/ESVD algorithm by processing a high-dimensional low rank matrix, which uses column pivoting QR decomposition to reduce dimension and ESVD to extract eigenvectors corresponding to nonzero eigenvalues. The experimental results on ORL, FERET and YALE face databases show that the proposed two algorithms can achieve almost the same performance and outperform the conventional DLDA algorithm in terms of computational complexity and training time. In addition, the experimental results on random data matrices show that the DLDA/QR-ESVD algorithm achieves better performance than the DLDA/ESVD algorithm by processing high-dimensional low rank matrices.
基金Supported partially by the National Natural Science Foundation of China under Grant Nos.60620160097,60472060 and 60473039.
文摘Principal Component Analysis （PCA） and Linear Discriminant Analysis （LDA） are two popular feature extraction techniques in statistical pattern recognition field. Due to small sample size problem LDA cannot be directly applied to appearance-based face recognition tasks. As a consequence, a lot of LDA-based facial feature extraction techniques are proposed to deal with the problem one after the other. Nullspace Method is one of the most effective methods among them. The Nullspace Method tries to find a set of discriminant vectors which maximize the between-class scatter in the null space of the within-class scatter matrix. The calculation of its discriminant vectors will involve performing singular value decomposition on a high-dimensional matrix. It is generally memory- and time-consuming. Borrowing the key idea in Nullspace method and the concept of coefficient of variance in statistical analysis we present a novel facial feature extraction method, i.e., Discriminant based on Coefficient of Variance （DCV） in this paper. Experimental results performed on the FERET and AR face image databases demonstrate that DCV is a promising technique in comparison with Eigenfaces, Nullspace Method, and other state-of-the-art facial feature extraction methods.
基金support of National Key Research and Development Program of China(2020YFA0908303)National Natural Science Foundation of China(21878081).
文摘Visual process monitoring is important in complex chemical processes.To address the high state separation of industrial data,we propose a new criterion for feature extraction called balanced multiple weighted linear discriminant analysis(BMWLDA).Then,we combine BMWLDA with self-organizing map(SOM)for visual monitoring of industrial operation processes.BMWLDA can extract the discriminative feature vectors from the original industrial data and maximally separate industrial operation states in the space spanned by these discriminative feature vectors.When the discriminative feature vectors are used as the input to SOM,the training result of SOM can differentiate industrial operation states clearly.This function improves the performance of visual monitoring.Continuous stirred tank reactor is used to verify that the class separation performance of BMWLDA is more effective than that of traditional linear discriminant analysis,approximate pairwise accuracy criterion,max–min distance analysis,maximum margin criterion,and local Fisher discriminant analysis.In addition,the method that combines BMWLDA with SOM can effectively perform visual process monitoring in real time.
文摘Capturing the distributed platform with remotely controlled compromised machines using botnet is extensively analyzed by various researchers.However,certain limitations need to be addressed efficiently.The provisioning of detection mechanism with learning approaches provides a better solution more broadly by saluting multi-objective constraints.The bots’patterns or features over the network have to be analyzed in both linear and non-linear manner.The linear and non-linear features are composed of high-level and low-level features.The collected features are maintained over the Bag of Features(BoF)where the most influencing features are collected and provided into the classifier model.Here,the linearity and non-linearity of the threat are evaluated with Support Vector Machine(SVM).Next,with the collected BoF,the redundant features are eliminated as it triggers overhead towards the predictor model.Finally,a novel Incoming data Redundancy Elimination-based learning model(RedE-L)is built to classify the network features to provide robustness towards BotNets detection.The simulation is carried out in MATLAB environment,and the evaluation of proposed RedE-L model is performed with various online accessible network traffic dataset(benchmark dataset).The proposed model intends to show better tradeoff compared to the existing approaches like conventional SVM,C4.5,RepTree and so on.Here,various metrics like Accuracy,detection rate,Mathews Correlation Coefficient(MCC),and some other statistical analysis are performed to show the proposed RedE-L model's reliability.The F1-measure is 99.98%,precision is 99.93%,Accuracy is 99.84%,TPR is 99.92%,TNR is 99.94%,FNR is 0.06 and FPR is 0.06 respectively.
文摘For the case where all multivariate normal parameters are known, we derive a new linear dimension reduction (LDR) method to determine a low-dimensional subspace that preserves or nearly preserves the original feature-space separation of the individual populations and the Bayes probability of misclassification. We also give necessary and sufficient conditions which provide the smallest reduced dimension that essentially retains the Bayes probability of misclassification from the original full-dimensional space in the reduced space. Moreover, our new LDR procedure requires no computationally expensive optimization procedure. Finally, for the case where parameters are unknown, we devise a LDR method based on our new theorem and compare our LDR method with three competing LDR methods using Monte Carlo simulations and a parametric bootstrap based on real data.
文摘The recognition of pathological voice is considered a difficult task for speech analysis.Moreover,otolaryngologists needed to rely on oral communication with patients to discover traces of voice pathologies like dysphonia that are caused by voice alteration of vocal folds and their accuracy is between 60%–70%.To enhance detection accuracy and reduce processing speed of dysphonia detection,a novel approach is proposed in this paper.We have leveraged Linear Discriminant Analysis(LDA)to train multiple Machine Learning(ML)models for dysphonia detection.Several ML models are utilized like Support Vector Machine(SVM),Logistic Regression,and K-nearest neighbor(K-NN)to predict the voice pathologies based on features like Mel-Frequency Cepstral Coefficients(MFCC),Fundamental Frequency(F0),Shimmer(%),Jitter(%),and Harmonic to Noise Ratio(HNR).The experiments were performed using Saarbrucken Voice Data-base(SVD)and a privately collected dataset.The K-fold cross-validation approach was incorporated to increase the robustness and stability of the ML models.According to the experimental results,our proposed approach has a 70%increase in processing speed over Principal Component Analysis(PCA)and performs remarkably well with a recognition accuracy of 95.24%on the SVD dataset surpassing the previous best accuracy of 82.37%.In the case of the private dataset,our proposed method achieved an accuracy rate of 93.37%.It can be an effective non-invasive method to detect dysphonia.
基金Supported by the National High Technology Research and Development Programme of China （ No. 2007AA01Z401 ） and the National Natural Science Foundation of China （No. 90718003, 60973027）.
文摘In order to achieve failure prediction without manual intervention for distributed systems, a novel failure feature analysis and extraction approach to automate failure prediction is proposed. Compared with the traditional methods which focus on building heuristic rules or models, the autonomic prediction approach analyzes the nonlinear correlation of failure features by recognizing failure patterns. Failure data are sorted according to the nonlinear correlation and failure signature is proposed for autonomic prediction. In addition, the Manifold Learning algorithm named supervised locally linear embedding is applied to achieve feature extraction. Based on the runtime monitoring of failure metrics, the experimental results indicate that the proposed method has better performance in terms of both correlation recognition precision and feature extraction quality and thus it can be used to design efficient autonomic failure prediction for distributed systems.
基金supported by the Program for New Century Excellent Talents in University of China,the NUST Outstanding Scholar Supporting Program,and the National Natural Science Foundation of China(Grant No.60973098).
文摘This paper introduces an idea of generating a kernel from an arbitrary function by embedding the training samples into the function.Based on this idea,we present two nonlinear feature extraction methods:generating kernel principal component analysis(GKPCA)and generating kernel Fisher discriminant(GKFD).These two methods are shown to be equivalent to the function-mapping-space PCA(FMS-PCA)and the function-mapping-space linear discriminant analysis(FMS-LDA)methods,respectively.This equivalence reveals that the generating kernel is actually determined by the corresponding function map.From the generating kernel point of view,we can classify the current kernel Fisher discriminant(KFD)algorithms into two categories:KPCA+LDA based algorithms and straightforward KFD(SKFD)algorithms.The KPCA+LDA based algorithms directly work on the given kernel and are not suitable for non-kernel functions,while the SKFD algorithms essentially work on the generating kernel from a given symmetric function and are therefore suitable for non-kernels as well as kernels.Finally,we outline the tensor-based feature extraction methods and discuss ways of extending tensor-based methods to their generating kernel versions.
文摘局部线性嵌入算法采用欧氏距离选择邻域点,这通常会损失数据集本身的非线性特征,造成邻域点选取错误,且仅使用欧氏距离构造权重会导致信息挖掘不充分。针对以上问题,提出基于概率模型与信息熵的局部线性嵌入算法(Probability information entropy-LLE,PIE-LLE)。首先,为了使邻域点选择更加合理,从数据集的概率分布角度出发,考虑样本点及其邻域的概率分布,为样本点构造符合局部分布的邻域集合。其次,为了充分提取样本的局部结构信息,在权重构造阶段,分别计算样本所属邻域概率以及每个样本的信息熵,融合二者信息重构低维样本。最后,在两个轴承故障数据集上的实验表明,所提方法故障识别准确度最高达到了100%,高于其他对比算法;在邻域点个数5~15范围内,PIE-LLE算法展现出良好的低维可视化效果;在参数敏感性实验中,该算法可以保持Fisher指标较大,有效提高了算法的分类准确度和稳定性。
