Discriminative Space Research Articles

Manifold Adaptive Kernel Semisupervised Discriminant Analysis for Gait Recognition

A manifold adaptive kernel semisupervised discriminant analysis algorithm for gait recognition is proposed in this paper. Motivated by the fact that the nonlinear structure captured by the data-independent kernels (such as Gaussian kernel, polynomial kernel, and Sigmoid kernel) may not be consistent with the discriminative information and the intrinsic manifold structure information of gait image, we construct two graph Laplacians by using the two nearest neighbor graphs (i.e., an intrinsic graph and a penalty graph) to model the discriminative manifold structure. We then incorporate these two graph Laplacians into the kernel deformation procedure, which leads to the discriminative manifold adaptive kernel space. Finally, the discrepancy-based semi-supervised discriminant analysis is performed in the manifold adaptive kernel space. Experimental results on the well-known USF HumanID gait database demonstrate the efficacy of our proposed algorithm.

Misclassification error propagation in land cover change categorization

It is important to describe misclassification errors in land cover maps and to quantify their propagation through geo-processing to resultant information products, such as land cover change maps. Geostatistical simulation is widely used in error modeling, as it can generate equal-probable realizations of the fields being considered, which can be summarized to facilitate error propagation analysis. To fix noninvariance in indicator simulation, discriminant space-based methods were proposed to enhance consistency in area-class mapping and replicability in uncertainty modeling, as the former is achieved by imposing means while the latter is ensured by projecting spatio-temporal correlated residuals in discriminant space to geographic space through a mapping process. This paper explores discriminant models for error propagation in land cover change detection, followed by experiments based on bi-temporal remote sensing images. It was found that misclassification error propagation is effectively characterized with discriminant covariate-based stochastic simulation, where spatio-temporal interdependence is taken into account.

Face recognition using various scales of discriminant color space transform

Research on color face recognition in the existing literature is aimed to establish a color space that can have the most of the discriminative information from the original data. This mainly includes optimal combination of different color components from the original color space. Recently proposed discriminate color space (DCS) is theoretically optimal for classification, in which one seeks a set of optimal coefficients in terms of linear combinations of the R, G and B components (based on a discriminate criterion). This work proposes an innovative block-wise DCS (BWDCS) method, which allows each block of the image to be in a distinct DCS. This is an interesting alternative to the methods relying on converting whole image to DCS. This idea is evaluated with four appearance-based subspace state-of-the-art methods on five different publicly available databases including the well-known FERET and FRGC databases. Experimental results show that the performance of these four gray-scale based methods can be improved by 17% on average when they are used with the proposed color space.

Geographic variation in trace-element signatures in the statoliths of near-hatch larvae and recruits of Concholepas concholepas (loco)

Spatial variation of trace elements in calcified structures (otoliths, statoliths, and shells) has been used to track the movements of individuals among habitats, and connectivity between marine populations. In the present study, we used laser ablation−inductively coupled plasma mass spectrometry to quantify the concentrations of trace elements in statoliths of prehatch larvae and recruits of the gastropod Concholepas concholepas from 3 regions in Chile. We also examined spatial variation in chemical signatures deposited during larval life and at the time of settlement in intertidal habitats. We found significant differences between 3 geographic regions in the trace element concentrations recorded in natal statoliths of near-hatch larvae and in natal core and edge areas of recruit statoliths. Discriminant function analysis indicates that natal signatures of near-hatch larvae and the cores and edges of recruit statoliths show spatial segregation among regions. High levels of reclassification success of larvae to the origin region suggest potential for assigning recruits to the corresponding matching region. Concentrations of trace elements in the natal cores of recruit statoliths fell relatively close but did not overlap with the discriminant space occupied by larvae, and at regional scales the pattern of geographic variation of recruit statoliths resembles that of larval statoliths. This suggests population grouping and little population interchange at this regional scale. Assessing population stocks and connectivity of this species at smaller scales along the Chilean coast will only be possible with more finely structured sampling and a better understanding of temporal variation in the chemical environment.

Uncertainty Comparison for Area-Class Maps Concerning Different Reference Data

As the main factor that influences classification quality, uncertainty characterization is analysis of area classes based on remotely sensed imagery and auxiliary data. This study focuses on uncertainty comparison between reference and classification maps. By referring to information classes and data classes respectively, experiment using real data sets was carried out to quantify uncertainty in area-class maps. Contingency tables and an information theory measure of shared information, percentage of average mutual information (%AMI), were applied to compare the uncertainty between pairs of area-class maps, where maximum likelihood classifier was used to classify the image into area-class map with different reference data on the discriminant space. Results show that there exist large impacts of semantic bias in different reference on classification uncertainty. Therefore, further improving upon the effect of bias in reference data will be studied to enable a more accurate assessment of the quality of classification.

Discriminant models for uncertainty characterization in area class change categorization

Discriminant space defining area classes is an important conceptual construct for uncertainty characterization in area-class maps. Discriminant models were promoted as they can enhance consistency in area-class mapping and replicability in error modeling. As area classes are rarely completely separable in empirically realized discriminant space, where class inseparability becomes more complicated for change categorization, we seek to quantify uncertainty in area classes (and change classes) due to measurement errors and semantic discrepancy separately and hence assess their relative margins objectively. Experiments using real datasets were carried out, and a Bayesian method was used to obtain change maps. We found that there are large differences between uncertainty statistics referring to data classes and information classes. Therefore, uncertainty characterization in change categorization should be based on discriminant modeling of measurement errors and semantic mismatch analysis, enabling quantification of uncertainty due to partially random measurement errors, and systematic categorical discrepancies, respectively.

楽曲に対する感性評価と音響ゆらぎ特徴の対応付け

This paper relates audio features based on fluctuation of music with sensibility evaluation about music as a component technology for an automatic song select system demanded on human sensibility evaluation. When people listen to music, they select a song considering both their own feelings at the time and sensibility evaluation about the song. We consider that sensibility evaluations of songs are influenced by the features based on fluctuation about both volume and pitch of the songs. Thus, we focus on features of fluctuation that contain a dynamic idea on music, and extract thirty six features of fluctuation about both volume and pitch from each songs using Fast Fourier Transform. On the other, we prepare a subjective experiment for plural songs using Semantic Differential method, and obtain the sensibility evaluation about each song. Then, we study the relationships between extracted features and sensibility evaluation about the songs with multiple discriminant analysis. As a result, high accurate discriminant hit-rates and low discriminant error are shown, therefore we suggest that audio features based on fluctuation of songs influence sensibility evaluation about the songs. Furthermore, we confirm the especial parameters related with sensibility evaluation about music while considering canonical variates which construct discriminant spaces.

Incremental linear discriminant analysis for evolving feature spaces in multitask pattern recognition problems

In this paper, we propose a new incremental linear discriminant analysis (ILDA) for multitask pattern recognition (MTPR) problems in which a chunk of training data for a particular task are given sequentially and the task is switched to another related task one after another. The Pang et al.’s ILDA is extended such that a discriminant space of the current task is augmented with effective discriminant vectors that are selected from other tasks based on the class separability. We call this selective augmentation of discriminant vectors knowledge transfer of feature space. In the experiments, the proposed ILDA is evaluated for seven MTPR problems, each of which consists of three recognition tasks. The results demonstrate that the proposed ILDA with knowledge transfer outperforms the conventional ILDA and its naive extension to MTPR problems with regard to both class separability and recognition accuracy. We confirm that the proposed knowledge transfer works well to evolve effective feature spaces online in MTPR problems.

Some statistical properties of differencing schemes for baseline correction of sedimentation velocity data

For the detailed analysis of sedimentation velocity data, the consideration of radial-dependent baseline offsets is indispensable. Two main approaches are data differencing (“delta-c” approach) and explicit inclusion of baseline parameters in the model (“direct boundary model” of the raw data). The current work aims to clarify the relationships between the two approaches. To this end, a simple model problem is examined. We show that the explicit consideration of the baseline in the model is equivalent to a differencing scheme where the average value is subtracted from all data points. Pairwise differencing in the delta-c approach always results in higher parameter uncertainty. For equidistant time points, the increase is smallest when the reference points are taken at intervals of 1/3 or 2/3 of the total number of time points. If the difference data are misinterpreted to be statistically independent samples, errors in the calculation of the parameter uncertainties can occur. Contrary to claims in the literature, we observe that there is no distinction in the approaches regarding their “model dependence”; both approaches arise from the integral or differential form of the same model, and both approaches can and should provide explicit estimates of the baseline values in the original data space for optimal discrimination between macromolecular sedimentation models.

Linking snake habitat use to nest predation risk in grassland birds: the dangers of shrub cover

Extremes in rangeland management, varying from too-frequent fire and intensive grazing to the suppression of both, threaten rangeland ecosystems worldwide. Intensive fire and grazing denude and homogenize vegetation whereas their suppression increases woody cover. Although habitat loss is implicated in grassland bird declines, degradation through intensive management or neglect also decreases breeding habitat and may reduce nesting success through increased rates of nest predation. Snakes are important nest predators, but little is known about how habitat use in snakes relates to predation risk for grassland birds nesting within tallgrass prairie subjected to different grazing and fire frequencies. We evaluated nest survival in the context of habitat used by nesting songbirds and two bird-eating snakes, the eastern yellowbelly racer Coluber constrictor flaviventris and Great Plains ratsnake Pantherophis emoryi. Daily nest survival rates decreased with increasing shrub cover and decreasing vegetation height, which characterize grasslands that have been neglected or intensively managed, respectively. Discriminant function analysis revealed that snake habitats were characterized by higher shrub cover, whereas successful nests were more likely to occur in areas with tall grass and forbs but reduced shrub cover. Because snakes often use shrub habitat, birds nesting in areas with increased shrub cover may be at higher risk of nest predation by snakes in addition to other predators known to use shrub habitat (e.g., mid-sized carnivores and avian predators). Depredated nests also occurred outside the discriminant space of the snakes, indicating that other predators (e.g., ground squirrels Spermophilus spp. and bullsnakes Pituophis catenifer) may be important in areas with denuded cover. Targeted removal of shrubs may increase nest success by minimizing the activity of nest predators attracted to shrub cover.

Projection-Based Location System via Multiple Discriminant Analysis in Wireless Local Area Networks

This paper presents a projection-based location system in an indoor wireless local area network (WLAN) environment. Our algorithm projects the received signal strength (RSS) onto a discriminative space such that the information of all access points (APs) is more efficiently utilized. The projection is determined by multiple discriminant analysis (MDA), thereby guaranteeing maximal discriminative information involved in the positioning system. The study conducts a series of experiments on the effects of our approach in a realistic indoor environment. The results show that not only is the positioning accuracy significantly improved, but the system cost, including the computation and data collection, is also greatly reduced at the same time. This is because our approach extracts only useful information for positioning, whereas the redundant noise is discarded to avoid the problem of overfitting and unnecessary calculations. Compared with prior works, this technique can produce a more graceful balance between the positioning accuracy and the computational complexity for the resource-weak clients.

(Post)bureaucratic organizational practices and the production of racioethnic inequality at work

AbstractExplanations of racioethnic inequality in organizations highlight the role of stereotypes, prejudices, biases, and discrimination in the production of such inequality. However, little is known about the impact that regular organizational practices have, either curbing or exacerbating racioethnic inequality at work. This paper argues that bureaucratic practices curb while post-bureaucratic practices exacerbate racioethnic inequality. First, this paper shows that unequal access to career advancements and payment between majority and minority employees of the Dutch national tax administration is partly due to the prevalence of post-bureaucratic ways of labour control. Second, linked to forms of labour control, post-bureaucratic ways of organizing the primary process also contribute to the production of such racioethnic inequality. Regarding both labour control and the structuring of the primary process, minority employees have an interest in the application of bureaucratic concepts that limit the space for stereotypes, prejudices, biases, and discrimination whereas post-bureaucratic concepts amplify this space.

(Post)bureaucratic organizational practices and the production of racioethnic inequality at work

AbstractExplanations of racioethnic inequality in organizations highlight the role of stereotypes, prejudices, biases, and discrimination in the production of such inequality. However, little is known about the impact that regular organizational practices have, either curbing or exacerbating racioethnic inequality at work. This paper argues that bureaucratic practices curb while post-bureaucratic practices exacerbate racioethnic inequality. First, this paper shows that unequal access to career advancements and payment between majority and minority employees of the Dutch national tax administration is partly due to the prevalence of post-bureaucratic ways of labour control. Second, linked to forms of labour control, post-bureaucratic ways of organizing the primary process also contribute to the production of such racioethnic inequality. Regarding both labour control and the structuring of the primary process, minority employees have an interest in the application of bureaucratic concepts that limit the space for stereotypes, prejudices, biases, and discrimination whereas post-bureaucratic concepts amplify this space.

Discriminant Models of Uncertainty in Nominal Fields

AbstractDespite developments in error modeling in discrete objects and continuous fields, there exist substantial and largely unsolved conceptual problems in the domain of nominal fields. This article explores a novel strategy for uncertainty characterization in spatial categorical information. The proposed strategy is based on discriminant space, which is defined with essential properties or driving processes underlying spatial class occurrences, leading to discriminant models of uncertainty in area classes. This strategy reinforces consistency in categorical mapping by imposing class‐specific mean structures that can be regressed against discriminant variables, and facilitates scale‐dependent error modeling that can effectively emulate the variation found between observers in terms of classes, boundary positions, numbers of polygons, and boundary network topology. Based on simulated data, comparisons with stochastic simulation based on indicator kriging confirmed the replicability of the discriminant models, which work by determining the mean area classes based on discriminant variables and projecting spatially correlated residuals in discriminant space to uncertainty in area classes.

An Illumination Invariant Bimodal Method Employing Discriminant Features for Face Recognition

A novel bimodal method for face recognition under low-level lighting conditions is proposed. It fuses an enhanced gray level imageand an illumination-invariant geometric image at the feature-level. To further improve the recognition performance under large variations in attributions such as poses and expressions, discriminant features are extracted from source images using the wavelet transform-based method. Features are adaptively fused to reconstruct the final face sample. Then FLD is used to generate a supervised discriminant space for the classification task. Experiments show that the bimodal method outperforms conventional methods under complex conditions.

Discriminant analysis based on modified generalised singular value decomposition and its numerical error analysis

Generalised singular value decomposition (GSVD) has been used in the literature for linear discriminant analysis (LDA) to solve the small sample size problem in pattern recognition. However, this method, commonly known as LDA/GSVD algorithm, suffers from excessive computational load when the sample dimension is high. Here the GSVD framework used in the LDA/GSVD algorithm is modified by replacing the SVD of a high-dimension matrix with the eigen-decomposition of a small size inner product matrix, thus circumventing the direct calculation of a high-dimension singular vector matrix. It is established by a theorem that if the samples are linearly independent in the feature space, the samples in each class are degenerated into a distinct single point of a discriminative space derived from the GSVD-based algorithms, and the distances between the points depend only on the respective numbers of the samples in the corresponding classes. In order to overcome the over-fitting problem, a method to orthogonalise the basis of the discriminative subspace is proposed. The proposed linear algorithm is kernelised for the discriminant analysis of samples that are not linearly independent as the non-linear kernel mapping can establish linear independence. The results of the above theorem are used to develop a method to measure the numerical error. This measure can also be used to decide the kernel parameters to minimise the numerical error in the non-linear algorithm.

Characteristics of grouping colors for figure segregation on a multicolored background

A figure is segregated from its background when the colored elements belonging to the figure are grouped together. We investigated the range of color distribution conditions in which a figure could be segregated from its background using the color distribution differences. The stimulus was a multicolored texture composed of randomly shaped pieces. It was divided into two regions: a test region and a background region. The pieces in these two regions had different color distributions in the OSA Uniform Color Space. In our experiments, the subject segregated the figure of the test region using two different procedures. Since the Euclidean distance in the OSA Uniform Color Space corresponds to perceived color difference, if segregation thresholds are determined by only color difference, the thresholds should be independent of position and direction in the color space. In the results, however, the thresholds did depend on position and direction in the OSA Uniform Color Space. This suggests that color difference is not the only factor in figure segregation by color. Moreover, the threshold dependence on position and direction is influenced by the distances in the cone-opponent space whose axes are normalized by discrimination thresholds, suggesting that figure segregation threshold is determined by similar factors in the cone-opponent space for color discrimination. The analysis of the results by categorical color naming suggests that categorical color perception may affect figure segregation only slightly.

THE EVOLUTION OF GENDER EARNINGS GAPS AND DISCRIMINATION IN URBAN CHINA, 1988–95

This paper analyzes the impact of market liberalization on gender earnings differentials and discrimination against women in urban China at the beginning of the 1990s. The observed stability in the overall gender earnings gap between 1988 and 1995 is shown to result from a complex set of evolutions across enterprises, earnings distributions, and time. Our results highlight the interplay of opposing forces, with economic reforms contributing to changes in managers’ behaviors in different dimensions. On the one hand, by bringing more competition, liberalization favored a reduction in discriminating behaviors in both urban collectives and foreign‐invested enterprises; on the other hand, by relaxing institutional rules, it led to a loosening of the government's egalitarian wage‐setting policies, leaving more space for discrimination in state‐owned enterprises.

Towards automatic skill evaluation: Detection and segmentation of robot-assisted surgical motions

This paper reports our progress in developing techniques for “parsing” raw motion data from a simple surgical task into a labeled sequence of surgical gestures. The ability to automatically detect and segment surgical motion can be useful in evaluating surgical skill, providing surgical training feedback, or documenting essential aspects of a procedure. If processed online, the information can be used to provide context-specific information or motion enhancements to the surgeon. However, in every case, the key step is to relate recorded motion data to a model of the procedure being performed.Robotic surgical systems such as the da Vinci system from Intuitive Surgical provide a rich source of motion and video data from surgical procedures. The application programming interface (API) of the da Vinci outputs 192 kinematics values at 10 Hz. Through a series of feature-processing steps, tailored to this task, the highly redundant features are projected to a compact and discriminative space. The resulting classifier is simple and effective.Cross-validation experiments show that the proposed approach can achieve accuracies higher than 90% when segmenting gestures in a 4-throw suturing task, for both expert and intermediate surgeons. These preliminary results suggest that gesture-specific features can be extracted to provide highly accurate surgical skill evaluation.

A View Independent Video-Based Face Recognition Method Using Posterior Probability in Kernel Fisher Discriminant Space

This paper presents a view independent video-based face recognition method using posterior probability in Kernel Fisher Discriminant (KFD) space. In practical environment, the view of faces changes dynamically. Robustness to view changes is required for video-based face recognition in practical environment. Since the view changes induce large non-linear variation, kernel-based methods are appropriate. We use KFD analysis to cope with non-linear variation. To classify image sequence, the posterior probability in KFD space is used. KFD analysis assumes that the distribution of each class in high dimensional space is Gaussian. This makes the computation of posterior probability in KFD space easy. The combination of KFD space and posterior probability of image sequence is the main contribution of the proposed method. The performance is evaluated by using two face databases. Effectiveness of the proposed method is shown by the comparison with the other feature spaces and classification methods.