Speech emotion recognition using semi-supervised discriminant analysis

Semi-supervised discriminant analysis SDA which uses a combination of multiple embedding graphs and kernel SDA KSDA are adopted in supervised speech emotion recognition.When the emotional factors of speech signal samples are preprocessed different categories of features including pitch zero-cross rate energy durance formant and Mel frequency cepstrum coefficient MFCC as well as their statistical parameters are extracted from the utterances of samples.In the dimensionality reduction stage before the feature vectors are sent into classifiers parameter-optimized SDA and KSDA are performed to reduce dimensionality.Experiments on the Berlin speech emotion database show that SDA for supervised speech emotion recognition outperforms some other state-of-the-art dimensionality reduction methods based on spectral graph learning such as linear discriminant analysis LDA locality preserving projections LPP marginal Fisher analysis MFA etc. when multi-class support vector machine SVM classifiers are used.Additionally KSDA can achieve better recognition performance based on kernelized data mapping compared with the above methods including SDA.

LOCAL CORRELATION DISCRIMINANT ANALYSIS AND ITS SEMI-SUPERVISED EXTENSION

Considering limitations of Linear Discriminant Analysis (LDA) and Marginal Fisher Analysis (MFA), a novel discriminant analysis called Local Correlation Discriminant Analysis (LCDA) is proposed in this paper. The main idea behind LCDA is to use more robust similarity measure, correlation metric, to measure the local similarity between image data. This results in better classifi-cation performance. In addition, to further improve the discriminant power of LCDA, we extend LCDA to semi-supervised case, which can make use of both labeled and unlabeled data to perform dis-criminant analysis. Extensive experimental results on ORL and AR face databases demonstrate that the proposed LCDA and its semi-supervised version are superior to Principal Component Analysis (PCA), LDA, CEA, and MFA.

Cobalt crust recognition based on kernel Fisher discriminant analysis and genetic algorithm in reverberation environment

Recognition of substrates in cobalt crust mining areas can improve mining efficiency.Aiming at the problem of unsatisfactory performance of using single feature to recognize the seabed material of the cobalt crust mining area,a method based on multiple-feature sets is proposed.Features of the target echoes are extracted by linear prediction method and wavelet analysis methods,and the linear prediction coefficient and linear prediction cepstrum coefficient are also extracted.Meanwhile,the characteristic matrices of modulus maxima,sub-band energy and multi-resolution singular spectrum entropy are obtained,respectively.The resulting features are subsequently compressed by kernel Fisher discriminant analysis(KFDA),the output features are selected using genetic algorithm(GA)to obtain optimal feature subsets,and recognition results of classifier are chosen as genetic fitness function.The advantages of this method are that it can describe the signal features more comprehensively and select the favorable features and remove the redundant features to the greatest extent.The experimental results show the better performance of the proposed method in comparison with only using KFDA or GA.

Kernel Model Applied in Kernel Direct Discriminant Analysis for the Recognition of Face with Nonlinear Variations

A kernel-based discriminant analysis method called kernel direct discriminant analysis is employed, which combines the merit of direct linear discriminant analysis with that of kernel trick. In order to demonstrate its better robustness to the complex and nonlinear variations of real face images, such as illumination, facial expression, scale and pose variations, experiments are carried out on the Olivetti Research Laboratory, Yale and self-built face databases. The results indicate that in contrast to kernel principal component analysis and kernel linear discriminant analysis, the method can achieve lower （7%） error rate using only a very small set of features. Furthermore, a new corrected kernel model is proposed to improve the recognition performance. Experimental results confirm its superiority （1% in terms of recognition rate） to other polynomial kernel models.

Face Recognition Using Kernel Discriminant Analysis

Linear discrimiant analysis (LDA) has been used in face recognition. But it is difficult to handle the high nonlinear problems, such as changes of large viewpoint and illumination. In order to overcome these problems, kernel discriminant analysis for face recognition is presented. This approach adopts the kernel functions to replace the dot products of nonlinear mapping in the high dimensional feature space, and then the nonlinear problem can be solved in the input space conveniently without explicit mapping. Two face databases are given.

Robust Classification through a Nonparametric Kernel Discriminant Analysis

The problem of classification in situations where the assumption of normality in the data is violated, and there are non-linear clustered structures in the dataset is addressed. A robust nonparametric kernel discriminant classification function, which is able to address this challenge, has been developed and the misclassification rates computed for various bandwidth matrices. A comparison with existing parametric classification functions such as the linear discriminant and quadratic discriminant is conducted to evaluate the performance of this classification function using simulated datasets. The results presented in this paper show good performance in terms of misclassification rates for the kernel discriminant classifier when the correct bandwidth is selected as compared to other identified existing classifiers. In this regard, the study recommends the use of the proposed kernel discriminant classification rule when one wishes to classify units into one of several categories or population groups where parametric classifiers might not be applicable.

Linear Discriminant Analysis and Kernel Vector Quantization for Mandarin Digits Recognition

Linear discriminant analysis and kernel vector quantization are integrated into vector quantization based speech recognition system for improving the recognition accuracy of Mandarin digits. These techniques increase the class separability and optimize the clustering procedure. Speaker-dependent (SD) and speaker-independent (SI) experiments are performed to evaluate the performance of the proposed method. The experiment results show that the proposed method is capable of reaching the word error rate of 3.76% in SD case and 6.60 % in SI case. Such a system can be suitable for being embedded in personal digital assistant(PDA), mobile phone and so on to perform voice controlling such as digit dialing, calculating, etc.

Near-Infrared Spectroscopy Coupled with Kernel Partial Least Squares-Discriminant Analysis for Rapid Screening Water Containing Malathion

Near-infrared spectroscopy coupled with kernel partial least squares-discriminant analysis was used to rapidly screen water containing malathion. In the wavenumber of 4348 cm-1 to 9091 cm-1, the overall correct classification rate of kernel partial least squares-discriminant analysis was 100% for training set, and 100% for test set, with the lowest concentration detected malathion residues in water being 1 μg·ml-1. Kernel partial least squares-discriminant analysis was able to have a good performance in classifying data in nonlinear systems. It was inferred that Near-infrared spectroscopy coupled with the kernel partial least squares-discriminant analysis had a potential in rapid screening other pesticide residues in water.

Classification of Stateless People through a Robust Nonparametric Kernel Discriminant Function

Statelessness is the absence of any Nationality. These include the Pemba, Shona, Galjeel, people of Burundi and Rwanda descent, and children born in Kenya to British Overseas Citizens after 1983. Frequently, they are not only undocumented but also often overlooked and not included in National Administrative Registers. Accordingly, find it hard to participate in Social and Economic Affairs. There has been a major push by UNHCR and international partners to “map” the size of stateless populations and their demographic profile, as well as causes, potential solutions and human rights situation. One of the requirements by the UNHCR in their push is for countries to find a potential solution to statelessness which starts with classifying/associating a person from these communities to a particular local community that is recognized in Kenya. This paper addresses this problem by adopting a Robust Nonparametric Kernel Discriminant function to correctly classify the stateless communities in Kenya and compare the performance of this method with the existing techniques through their classification rates. This is because Non-parametric functions have proven to be more robust and useful especially when there exists auxiliary information which can be used to increase precision. The findings from this paper indicate that Nonparametric discriminant classifiers provide a good classification method for classifying the stateless communities in Kenya. This is because they exhibit lower classification rates compared to the parametric methods such as Linear and Quadratic discriminant functions. In addition, the finding shows that based on certain similarities in characteristics that exist in these communities that surround the Pemba Community, the Pemba community can be classified as Giriama or Rabai in which they seem to have a strong link. In this regard, the study recommends the use of the Kernel discriminant classifiers in classifying the stateless persons and that the Government of Kenya consider integrating/recognizing the Pemba community into Giriama or Rabai so that they can be issued with the National Identification Cards and be recognized as Kenyans.

Ensemble enhanced active learning mixture discriminant analysis model and its application for semi-supervised fault classification

As an indispensable part of process monitoring, the performance of fault classification relies heavily on the sufficiency of process knowledge. However, data labels are always difficult to acquire because of the limited sampling condition or expensive laboratory analysis, which may lead to deterioration of classification performance.To handle this dilemma, a new semi-supervised fault classification strategy is performed in which enhanced active learning is employed to evaluate the value of each unlabeled sample with respect to a specific labeled dataset.Unlabeled samples with large values will serve as supplementary information for the training dataset. In addition,we introduce several reasonable indexes and criteria, and thus human labeling interference is greatly reduced. Finally,the fault classification effectiveness of the proposed method is evaluated using a numerical example and the Tennessee Eastman process.

基于支持向量的Kernel判别分析

提出了一种新的基于支持向量的核化判别分析方法(SV-KFD).首先深入地分析了支持向量机(SVM)以及核化费舍尔判别分析(KernelFisher)方法的相互关系.基于作者证明的SVM本身所固有的零空间性质:SVM分类面的法向量在基于支持向量的类内散度矩阵条件下,具有零空间特性,提出了利用SVM的法向量定义核化的决策边界特征矩阵(KernelizedDecisionBoundaryFeatureMatrix,KDBFM)的方法.进一步结合均值向量的差向量构建扩展决策边界特征矩阵(Ex-KDBFM).最后以支持向量为训练集合,结合零空间方法来计算投影空间,该投影空间被用来从原始图像中提取判别特征.以人脸识别为例,作者在FERET和CAS-PEAL-R1大规模人脸图像数据库上对所提出的方法进行了实验验证,测试结果表明该方法具有比传统核判别分析方法更好的识别性能.

基于LDA+kernel-KNNFLC的语音情感识别方法

结合K近邻、核学习方法、特征线重心法和LDA算法,提出了用于情感识别的LDA+kernel-KNNFLC方法.首先针对先验样本特征造成的计算量庞大问题,采用重心准则学习样本距离,改进了核学习的K近邻方法;然后加入LDA对情感特征向量进行优化,在避免维度冗余的情况下,更好地保证了情感信息识别的稳定性.最后,通过对特征空间再学习,结合LDA的kernel-KNNFLC方法优化了情感特征向量的类间区分度,适合于语音情感识别.对包含120维全局统计特征的语音情感数据库进行仿真实验,对降维方案、情感分类器和维度参数进行了多组对比分析.结果表明,LDA+kernel-KNNFLC方法在同等条件下性能提升效果最显著.

基于SRKDA的系统故障演化过程分解方法研究

为研究系统故障演化过程中可能蕴含的多种演化特征,对演化过程的分解进行研究,提出基于谱回归核判别分析(SRKDA)的演化过程分解方法。首先介绍演化过程的特点和分解原理,其次论证对象集合对演化过程的可表示性,给出分解方法流程,最后进行实例分析。研究结果表明:分解演化过程本质上是对象与系统功能状态对应关系的确定,各对象集合都对应了各自的子演化过程;线性和非线性条件下对象可表示各种功能状态;对象标签矩阵须满足标签值的均匀分布特征;使用SRKDA算法可以确定最大准确度和最优对象标签集合,实现演化过程的分解;实例分析得到在20000次迭代后最大准确度为0.85,3个子演化过程分别包含41,33,26个对象。研究结果可为系统故障过程的特征分析提供参考方法。

A Novel Systematic Method of Quality Monitoring and Prediction Based on FDA and Kernel Regression

A novel systematic quality monitoring and prediction method based on Fisher discriminant analysis (FDA) and kernel regression is proposed. The FDA method is first used for quality monitoring. If the process is un-der normal condition, then kernel regression is further used for quality prediction and estimation. If faults have oc-curred, the contribution plot in the fault feature direction is used for fault diagnosis. The proposed method can ef-fectively detect the fault and has better ability to predict the response variables than principle component regression (PCR) and partial least squares (PLS). Application results to the industrial fluid catalytic cracking unit (FCCU) show the effectiveness of the proposed method.

Application of Kernel GDA to Performance Monitoring and Fault Diagnosis for Rotating Machinery

Faults in rotating machine are difficult to detect and identify,especially when the system is complex and nonlinear.In order to solve this problem,a novel performance monitoring and fault diagnosis method based on kernel generalized discriminant analysis(kernel GDA,KGDA)was proposed.Through KGDA,the data were mapped from the original space to the high-dimensional feature space.Then the statistic distance between normal data and test data was constructed to detect whether a fault was occurring.If a fault had occurred,similar analysis was used to identify the type of faults.The effectiveness of the proposed method was evaluated by simulation results of vibration signal fault dataset in the rotating machinery,which was scalable to different rotating machinery.

近红外高光谱成像技术结合偏最小二乘-判别分析所建模型快速鉴定核桃仁的品质

将近红外高光谱成像技术与偏最小二乘-判别分析(PLS-DA)结合,建立了快速无损鉴定核桃仁品质的分类模型。在900~1 700 nm全波长范围内,采集不同品质核桃仁的光谱数据,以平均光谱作为原始光谱,以标准正态变量对原始光谱数据进行预处理,采用主成分分析对原始光谱数据降维,提取到970,1 151,1 210,1 215,1 256,1 309,1 340,1 379,1 389,1 404,1 460 nm等11个特征波长。基于全光谱和特征波长,分别建立两种PLS-DA分类模型。结果表明:全光谱条件下所建模型在校准集和验证集上的预测正确率最高,可达100%;特征波长条件下所建模型在相同数据集上的分类正确率略有下降,为99.3%;两种模型在测试集上的预测正确率均为100%。

基于核熵成分分析的工业过程多类型故障诊断

核熵成分分析(KECA)特征提取过程中只保留了数据的最大瑞丽熵(Renyi)信息,没有充分利用数据的类别信息。由于监督学习算法线性判别分析(LDA)能够有效提取特征中的类别信息,因此提出KECA-LDA(KEDA)的特征提取方法。首先KECA依据最小Renyi熵损失策略对数据进行维数约简;然后在KECA特征空间使用LDA算法获取具有判别信息的低维特征并输入到支持向量机(SVM)分类器中,利用天牛须优化算法(BAS)得到最佳性能的SVM分类器,从而建立故障诊断模型。将KEDA-BAS-SVM方法应用于田纳西-伊斯曼化工过程(TE)进行仿真实验,结果表明:当采用基于距离测度的矩阵相似性优化确定KEDA中所选用的径向基函数(RBF)核参数时,相比KECA和LDA算法,KEDA特征提取后多类型故障诊断准确率达到99.7%,验证了KEDA-BAS-SVM在多类型故障诊断领域的优越性。

基于核Fisher判别分析的高光谱遥感影像分类

高光谱遥感技术,将反映目标辐射特性的光谱信息与反映目标空间位置关系的图像信息有机地结合在一起。高光谱影像具有丰富的光谱信息,较全色、多光谱影像能够更好的进行地面目标的分类识别。在介绍核Fisher判别分析算法的基础上,选用径向基核函数,使用一对一或一对余构造多类构造法,并利用交叉验证网格搜索法优化核函数参数,构建了快速稳定的多类核Fisher判别分析分类器。通过OMIS和AVIRIS影像的分类实验,表明了核Fisher判别分析与支持向量机的分类精度相当,但是所需的训练时间较短。

一种基于核的快速非线性鉴别分析方法

基于"核技巧"提出的新的非线性鉴别分析方法在最小二乘意义上与基于核的Fisher鉴别分析方法等效,相应鉴别方向通过一个线性方程组得出,计算代价较小,相应分类实现极其简便.该方法的最大优点是,对训练数据进行筛选,可使构造鉴别矢量的"显著"训练模式数大大低于总训练模式数,从而使得测试集的分类非常高效;同时,设计出专门的优化算法以加速"显著"训练模式的选取.实验表明,这种非线性方法不仅具有明显的效率上的优势,且具有不低于基于核的Fisher鉴别分析方法的性能.

基于广义判别分析的光谱分类

提出了基于广义判别分析(generalized discriminant analysis,GDA)方法对恒星(Star)、星系(Galaxy)和类星体(Quasars)的光谱进行分类。广义判别分析将核技巧与Fisher判别分析结合起来,通过非线性映射将样本集映射到高维特征空间F,在F空间中进行线性判别分析。实验对比了LDA,GDA,PCA,KPCA算法对于恒星、星系和类星体的光谱分类性能。结果表明基于GDA的算法对于这3种类型光谱的分类正确率最高,LDA次之;尽管KPCA也是一种基于核的方法,但是选择主成分个数较少时效果较差,甚至低于LDA;基于PCA的分类效果最差。