Spectral slope based analysis and classification of stressed speech

In this study, spectral slope based features are investigated for characterization and classification of stressed speech. The vocal tract spectrum is modulated with glottal flow spectra, resulting a tilt in the overall spectrum. In this study, spectral tilt is analyzed for different stress classes. Relative formant peak displacement (RFD) is proposed as the displacement of formant peaks from the 1 ^st formant peak. The displacement of 2 ^nd , 3 ^rd and 4 ^th formant peaks from 1 ^st formant peak is termed as RFD ₂, RFD ₃ and RFD ₄, respectively. The features are extracted from linear prediction coefficient (LPC) and cepstrally smoothed log spectrum, respectively. Analysis shows that stress effects higher formant region more than lower formant region. To evaluate the effectiveness of this feature for different stress classes, the performance of stress classification is evaluated. A simulated stressed speech database is collected under four stress conditions, namely, neutral, angry, sad and Lombard from fifteen speakers. The performance of RFD feature is similar to Mel-frequency cepstral coefficient (MFCC). This shows that RFD feature have approximately same discrimination capability for stress as MFCC. Further, the performance of cepstrally smoothed log spectra derived RFD are higher than LPC derived RFD feature. RFD features are combined with MFCC in feature, score and rank level and found improved performance.     

Spectral analysis of sensor data in civil engineering structures

A system identification procedure is developed for processing structural response signals registered by sensors permanently installed for monitoring at critical locations of a structure. The procedure, which is based on spectrogram estimation and is applied within a sweeping window of signal samples, was built into a data processing software that allows the determination of the fundamental frequencies of vibrating structural members and can be used to identify the changing properties of instrumented structural systems through time. Field data obtained from strain sensors on a specialized bridge deck structural system that spans the Salmon River in Nova Scotia, Canada were used as case study for evaluating the performance of the proposed method. Successful correlation was obtained between theoretical and observed vibration characteristics, whereas several important features of the actual function of various bridge members were revealed through the analysis procedure.     

医院临床数据分析智能分类处理技术研究

针对医院临床数据数量庞大,数据之间关联性大,容易出现数据提取不准确等问题,提出基于模糊分类处理技术的医院临床数据智能分类方法。通过对临床运营各项指标的说明,根据指标分析数据的特性;对医院临床数据进行检索,将检索出来的数据进行提取,根据数据的特点,使用模糊分类的技术对数据进行处理,完成临床数据的智能分类。实验结果表明,所提方法对临床数据的分类效果远远优于传统方法,满足了医院对数据处理的要求,为未来医院大量的数据分类处理奠定了坚实的基础。     

Spectral principal component analysis of dynamic process data

This article describes principal component analysis (PCA) of the power spectra of data from chemical processes. Spectral PCA can be applied to the measurements from a whole unit or plant because spectra are invariant to the phase lags caused by time delays and process dynamics. The same comment applies to PCA using autocovariance functions, which was also studied. Two case studies are presented. One was derived from simulation of a pulp process. The second was from a refinery involving 37 tags. In both cases, PCA clusters were observed which were characterised by distinct spectral features. Spectral PCA was compared with PCA using autocovariance functions. The performance was similar, and both offered an improvement over PCA using the time domain signals even when time shifting was used to align the phases.     

Spectral analysis of irregularly-sampled data: Paralleling the regularly-sampled data approaches

The spectral analysis of regularly-sampled (RS) data is a well-established topic, and many useful methods are available for performing it under different sets of conditions. The same cannot be said about the spectral analysis of irregularly-sampled (IS) data: despite a plethora of published works on this topic, the choice of a spectral analysis method for IS data is essentially limited, on either technical or computational grounds, to the periodogram and its variations. In our opinion this situation is far from satisfactory, given the importance of the spectral analysis of IS data for a considerable number of applications in such diverse fields as engineering, biomedicine, economics, astronomy, seismology, and physics, to name a few.In this paper we introduce a number of IS data approaches that parallel the methods most commonly used for spectral analysis of RS data: the periodogram (PER), the Capon method (CAP), the multiple-signal characterization method (MUSIC), and the estimation of signal parameters via rotational invariance technique (ESPRIT). The proposed IS methods are as simple as their RS counterparts, both conceptually and computationally. In particular, the fast algorithms derived for the implementation of the RS data methods can be used mutatis mutandis to implement the proposed parallel IS methods. Moreover, the expected performance-based ranking of the IS methods is the same as that of the parallel RS methods: all of them perform similarly on data consisting of well-separated sinusoids in noise, MUSIC and ESPRIT outperform the other methods in the case of closely-spaced sinusoids in white noise, and CAP outperforms PER for data whose spectrum has a small-to-medium dynamic range (MUSIC and ESPRIT should not be used in the latter case).     

An improved genetic-fuzzy system for classification and data analysis

Interpretability of classification systems, which refers to the ability of these systems to express their behavior in an understandable way, has recently gained more attention and it is considered as an important requirement especially for knowledge-based systems. The main objective of this study is to improve the ability of a well-known fuzzy classifier proposed in Ishibuchi and Nojima (2007) to maximize the accuracy while preserve its interpretability. To achieve the above-mentioned objective, we propose two variants of the original fuzzy classifier. In the first variant classifier, the same components of the original classifier were used except NSGA-II which was replaced by an enhanced version called Controlled Elitism NSGA-II. This replacement aims at improving the ability of the first variant classifier to find non-dominated solutions with better interpretability-accuracy trade-off. In the second variant classifier, we further improve the first variant classifier by enhancing the selection method of the antecedent conditions of the rules generated in the initial population of genetic algorithm. Unlike the method applied in the original classifier and the first variant classifier, which uses a random selection of the antecedent conditions, we proposed a feature-based selection method to favor the antecedent conditions associated with the most relevant features. The results show that the two variant classifiers find more non-dominated fuzzy rule-based systems with better generalization ability than the original method which suggests that Controlled Elitism NSGA-II algorithm is more efficient than NSGA-II. In addition, feature-based selection method applied in the second variant classifier allowed this method to successfully obtain high-quality solutions as it has consistently achieved the best error rates for all the data sets compared to the original method and the first variant classifier.     

基于分布式数据仓库的分类分析研究

针对GAC-RDB分类算法只能应用于单机版数据仓库的局限性,为了能够更方便、快捷地在云计算平台上开展数据挖掘工作,基于分布式数据仓库HBase,结合GAC-RDB分类算法的实现机理,制定适合分布式平台的运行策略,使用原生HiveQL语言提出了一种分布式GAC-RDB分类算法。实验显示,随着集群中节点的不断增加,算法的运行时间稳步下降。结果表明,在保证算法准确率的前提下,分布式数据仓库能够有效提高GACRDB分类算法的扩展性和运行效率,相对于MapReduce框架,HiveQL语言降低了对数据挖掘从业人员的技术要求,更大程度地减少了算法的开发时间,为挖掘海量数据提供了新的解决方案。     

Error entropy in classification problems: a univariate data analysis

Entropy-based cost functions are enjoying a growing attractiveness in unsupervised and supervised classification tasks. Better performances in terms both of error rate and speed of convergence have been reported. In this letter, we study the principle of error entropy minimization (EEM) from a theoretical point of view. We use Shannon's entropy and study univariate data splitting in two-class problems. In this setting, the error variable is a discrete random variable, leading to a not too complicated mathematical analysis of the error entropy. We start by showing that for uniformly distributed data, there is equivalence between the EEM split and the optimal classifier. In a more general setting, we prove the necessary conditions for this equivalence and show the existence of class configurations where the optimal classifier corresponds to maximum error entropy. The presented theoretical results provide practical guidelines that are illustrated with a set of experiments with both real and simulated data sets, where the effectiveness of EEM is compared with the usual mean square error minimization.     

Spectral analysis of nonuniformly spaced data using least square method

The current study investigates an unevenly spaced spectrum using a least square method. The well known Lomb periodogram approach has many benefits, yet it cannot resolve positive and negative frequencies. Using a modified scheme, positive and negative frequencies were discerned without losing any of the benefits of a Lomb periodogram. One of the properties of the periodogram approach is the relationship between the coefficients, i.e., the Hilbert transformation pair. By utilizing this property, the processing time was reduced by half.     

Spectral feature selection for shape characterization and classification

This paper proposes a framework for selecting the Laplacian eigenvalues of 3D shapes that are more relevant for shape characterization and classification. We demonstrate the redundancy of the information coded by the shape spectrum and discuss the shape characterization capability of the selected eigenvalues. The feature selection methods used to demonstrate our claim are the AdaBoost algorithm and Support Vector Machine. The efficacy of the selection is shown by comparing the results of the selected eigenvalues on shape characterization and classification with those related to the first k eigenvalues, by varying k over the cardinality of the spectrum. Our experiments, which have been performed on 3D objects represented either as triangle meshes or point clouds, show that working directly with point clouds provides classification results that are comparable with respect to those related to surface-based representations. Finally, we discuss the stability of the computation of the Laplacian spectrum to matrix perturbations.     

Increasing the efficiency of combinatorial logical data analysis in recognition and classification problems

Problems of increasing the efficiency of combinatorial logical data analysis in recognition problems are examined. A technique for correct conversion of initial information for reduction of its dimensionality is proposed. Results of testing this technique for problems of real medical prognoses are given. Djukova Elena V. Born 1945. Graduated from Moscow State University in 1967. Candidate’s degree in Physics and Mathematics in 1979. Doctoral degree in Physics and Mathematics in 1997. Dorodnicyn Computing Center, Russian Academy of Sciences, leading researcher. Moscow State University, lecturer. Moscow Pedagogical University, lecturer. Scientific interests: discrete mathematics and mathematical method of pattern recognition. Author of 70 papers. Peskov Nikolai V. Born 1978. Graduated from Moscow State University in 2000. Candidate’s degree in 2004. Dorodnicyn Computing Center, Russian Academy of Sciences, junior researcher. Scientific interests: discrete mathematics and mathematical methods of pattern recognition. Author of ten papers. Inyakin Andrey S. Born 1978. Graduated from Moscow State University in 2000. Dorodnicyn Computing Center, Russian Academy of Sciences, junior researcher. Scientific interests: discrete mathematics and mathematical methods of pattern recognition. Author of ten papers. Sakharov Aleksei A. Born 1980. Graduated from Moscow State University in 2003. Moscow Pedagogical University, graduate student. Scientific interests: discrete mathematics and mathematical method of pattern recognition. Author of three papers.     

Evaluation and classification of otoneurological data with new data analysis methods based on machine learning

We improved the classification ability of multilayer perceptron networks by constructing a set of networks of as many as output classes and investigated the influence of different input variables on the classification. We have developed methods named scattering, spectrum and response analysis to express the classification complexity, especially the overlap of output classes, to disentangle the relation between the input variables and output classes of perceptron neural networks, and to establish the importance of input variables. The methods were tested by exploring complicated otoneurological data. In contrast to the variable selection problem, our methods characterize the importance of variables for classification and also describe the importance of the different values of each variable for output (disease) classes. When complex data is distributed in a biased manner between disease classes, we improved classification accuracy by developing a network set called NetSet, which increased average sensitivity and positive predictive value for at least 10% up to 85% and 83% respectively, compared to our earlier neural network classifications with the same data, which clarified class distribution effects and supported our comprehension of the significance of input.     

Incremental linear discriminant analysis for classification of data streams.

This paper presents a constructive method for deriving an updated discriminant eigenspace for classification when bursts of data that contains new classes is being added to an initial discriminant eigenspace in the form of random chunks. Basically, we propose an incremental linear discriminant analysis (ILDA) in its two forms: a sequential ILDA and a Chunk ILDA. In experiments, we have tested ILDA using datasets with a small number of classes and small-dimensional features, as well as datasets with a large number of classes and large-dimensional features. We have compared the proposed ILDA against the traditional batch LDA in terms of discriminability, execution time and memory usage with the increasing volume of data addition. The results show that the proposed ILDA can effectively evolve a discriminant eigenspace over a fast and large data stream, and extract features with superior discriminability in classification, when compared with other methods.     

Spectral characterization and regression-based classification of forest damage in Norway spruce stands in the Czech Republic using Landsat Thematic Mapper data

This study assessed the ability of Landsat Thematic Mapper (TM) sensor data to discriminate among three damage categories of Norway spruce in the Krusne Hory mountains using dichotomous logit regressions. Moderate and light damage stands, being the most spectrally similar, were separated with 83 per cent accuracy using TM1, TM4 and TM7. Moderate and heavy categories were best separated by TM3 (accuracy=88 per cent). Light and heavy damage classes were separated with up to 95 per cent accuracy. Ratios and indices did not improve the regression accuracies. The regression equations, when used to classify three categories of damage, accurately classified 71–75 per cent of Norway spruce stands.     

Joint learning of cross-modal classifier and factor analysis for multimedia data classification

In this paper, we study the problem of learning from multiple model data for the purpose of document classification. In this problem, each document is composed of two different models of data, i.e., an image and a text. We propose to represent the data of two models by projecting them to a shared data space by using cross-model factor analysis formula and classify them in the shared space by using a linear class label predictor, named cross-model classifier. The parameters of both cross-model classifier and cross-model factor analysis are learned jointly, so that they can regularize the learning of each other. We construct a unified objective function for this learning problem. With this objective function, we minimize the distance between the projections of image and text of the same document, and the classification error of the projections measured by hinge loss function. The objective function is optimized by an alternate optimization strategy in an iterative algorithm. Experiments in two different multiple model document data sets show the advantage of the proposed algorithm over state-of-the-art multimedia data classification methods.     

Mapping coffee crops in southeastern Brazil using spectral mixture analysis and data mining classification

The task of mapping coffee crops using multispectral data sets is not yet a trivial routine. This is because coffee fields are extremely heterogeneous in terms of spectral reflectance. This study therefore aims to contribute to the mapping of coffee crops using multispectral imagery with 23.5 m spatial resolution taken by the Linear Imaging Self Scanner (LISS III) instrument on board the Indian Remote Sensing (IRS) satellite system. The section of land covered by this study is a traditional coffee-producing province located in the south of the State of Minas Gerais, southeastern Brazil. Whereas the pixel mixture effect was managed using spectral mixture analysis (SMA), the classification was carried out using data mining (DM) techniques. The decision tree (DT) outcomes were evaluated using a simple and qualitative method based on the elements of photointerpretation. In total, eight land-use and land-cover (LULC) types were mapped, including three classes of coffee-growing land expressing different phenological conditions and management. These were named ‘Production Coffee’, ‘Mixed Coffee’, and ‘Old/Pruned Coffee’. The results showed that the methodology was effective for mapping LULC types, as the workflow adopted simplified image interpretation and offered improvements in the classification performance. Despite the coffee-cultivation classes having a large spectral variability, which increases the chances of classification errors, not many confusions were observed involving the three coffee classes mapped with other categories of use. This therefore shows that the method was efficient in isolating the coffee classes (with an accuracy greater than 70%) from other categories of use. Comparing the results obtained in this work with a conventional maximum-likelihood (ML) classification, the results revealed that when using the methodology described, the confusions between classes were less dispersed and an improvement of approximately 10% was observed in the mapping of the Production Coffee class.     

Spectral and temporal linear mixing model for vegetation classification

The objective of this study is to find a better method for sub-pixel classification of vegetation. The proposed new technique of a linear mixing model (LMM) is the sequential combination of spectral LMM and temporal LMM. Sub-pixel components of ‘relative green vegetation’ are derived by spectral LMM; sub-pixel components of vegetation types are estimated by subsequent temporal LMM. The proposed method was applied to five temporal Landsat Enhanced Thematic Mapper (ETM) images for the year 2000 for areas south of Lake Baikal, Russia. Dominant vegetation types there are pine, birch/aspen, shrubs and wheat with weedy plants. Ground truth data of vegetation types were prepared by field survey and visual interpretation of Landsat ETM images by experts. Both the comparisons of classification results among the proposed method and conventional LMM methods and the simulation results among them indicate that the proposed spectral and temporal LMM has better accuracy than conventional methods.