Applying SVMs and weight-based factor analysis to unsupervised adaptation for speaker verification

This paper presents an extended study on the implementation of support vector machine(SVM) based speaker verification in systems that employ continuous progressive model adaptation using the weight-based factor analysis model. The weight-based factor analysis model compensates for session variations in unsupervised scenarios by incorporating trial confidence measures in the general statistics used in the inter-session variability modelling process. Employing weight-based factor analysis in Gaussian mixture models (GMM) was recently found to provide significant performance gains to unsupervised classification. Further improvements in performance were found through the integration of SVM-based classification in the system by means of GMM supervectors. This study focuses particularly on the way in which a client is represented in the SVM kernel space using single and multiple target supervectors. Experimental results indicate that training client SVMs using a single target supervector maximises performance while exhibiting a certain robustness to the inclusion of impostor training data in the model. Furthermore, the inclusion of low-scoring target trials in the adaptation process is investigated where they were found to significantly aid performance.

Evaluation of spectral and texture features for object-based vegetation species classification using Support Vector Machines

The use of appropriate features to characterize an output class or object is critical for all classification problems. This paper evaluates the capability of several spectral and texture features for object-based vegetation classification at the species level using airborne high resolution multispectral imagery. Image-objects as the basic classification unit were generated through image segmentation. Statistical moments extracted from original spectral bands and vegetation index image are used as feature descriptors for image objects (i.e. tree crowns). Several state-of-art texture descriptors such as Gray-Level Co-Occurrence Matrix (GLCM), Local Binary Patterns (LBP) and its extensions are also extracted for comparison purpose. Support Vector Machine (SVM) is employed for classification in the object-feature space. The experimental results showed that incorporating spectral vegetation indices can improve the classification accuracy and obtained better results than in original spectral bands, and using moments of Ratio Vegetation Index obtained the highest average classification accuracy in our experiment. The experiments also indicate that the spectral moment features also outperform or can at least compare with the state-of-art texture descriptors in terms of classification accuracy.

Color and texture feature fusion using kernel PCA with application to object-based vegetation species classification

A good object representation or object descriptor is one of the key issues in object based image analysis. To effectively fuse color and texture as a unified descriptor at object level, this paper presents a novel method for feature fusion. Color histogram and the uniform local binary patterns are extracted from arbitrary-shaped image-objects, and kernel principal component analysis (kernel PCA) is employed to find nonlinear relationships of the extracted color and texture features. The maximum likelihood approach is used to estimate the intrinsic dimensionality, which is then used as a criterion for automatic selection of optimal feature set from the fused feature. The proposed method is evaluated using SVM as the benchmark classifier and is applied to object-based vegetation species classification using high spatial resolution aerial imagery. Experimental results demonstrate that great improvement can be achieved by using proposed feature fusion method.

Effects of reduced contrast on the perception and control of speed when driving

Misperception of speed under low-contrast conditions has been identified as a possible contributor to motor vehicle crashes in fog. To test this hypothesis, we investigated the effects of reduced contrast on drivers’ perception and control of speed while driving under real-world conditions. Fourteen participants drove around a 2.85 km closed road course under three visual conditions: clear view and with two levels of reduced contrast created by diffusing filters on the windscreen and side windows. Three dependent measures were obtained, without view of the speedometer, on separate laps around the road course: verbal estimates of speed; adjustment of speed to instructed levels (25 to 70 km h-1); and estimation of minimum stopping distance. The results showed that drivers traveled more slowly under low-contrast conditions. Reduced contrast had little or no effect on either verbal judgments of speed or estimates of minimum stopping distance. Speed adjustments were significantly slower under low-contrast than clear conditions, indicating that, contrary to studies of object motion, drivers perceived themselves to be traveling faster under conditions of reduced contrast. Under real-world driving conditions, drivers’ ability to perceive and control their speed was not adversely affected by large variations in the contrast of their surroundings. These findings suggest that perceptions of self-motion and object motion involve neural processes that are differentially affected by variations in stimulus contrast as encountered in fog.

Segmentation of scenes of mobile objects and demonstrable backgrounds

In this paper we present a real-time foreground–background segmentation algorithm that exploits the following observation (very often satisfied by a static camera positioned high in its environment). If a blob moves on a pixel p that had not changed its colour significantly for a few frames, then p was probably part of the background when its colour was static. With this information we are able to update differentially pixels believed to be background. This work is relevant to autonomous minirobots, as they often navigate in buildings where smart surveillance cameras could communicate wirelessly with them. A by-product of the proposed system is a mask of the image regions which are demonstrably background. Statistically significant tests show that the proposed method has a better precision and recall rates than the state of the art foreground/background segmentation algorithm of the OpenCV computer vision library.

A biologically inspired object spectral-texture descriptor and its application to vegetation classification in power-line corridors

The use of appropriate features to represent an output class or object is critical for all classification problems. In this paper, we propose a biologically inspired object descriptor to represent the spectral-texture patterns of image-objects. The proposed feature descriptor is generated from the pulse spectral frequencies (PSF) of a pulse coupled neural network (PCNN), which is invariant to rotation, translation and small scale changes. The proposed method is first evaluated in a rotation and scale invariant texture classification using USC-SIPI texture database. It is further evaluated in an application of vegetation species classification in power line corridor monitoring using airborne multi-spectral aerial imagery. The results from the two experiments demonstrate that the PSF feature is effective to represent spectral-texture patterns of objects and it shows better results than classic color histogram and texture features.

Multi-step ahead direct prediction for the machine condition prognosis using regression trees and neuro-fuzzy systems

This paper presents an approach to predict the operating conditions of machine based on classification and regression trees (CART) and adaptive neuro-fuzzy inference system (ANFIS) in association with direct prediction strategy for multi-step ahead prediction of time series techniques. In this study, the number of available observations and the number of predicted steps are initially determined by using false nearest neighbor method and auto mutual information technique, respectively. These values are subsequently utilized as inputs for prediction models to forecast the future values of the machines’ operating conditions. The performance of the proposed approach is then evaluated by using real trending data of low methane compressor. A comparative study of the predicted results obtained from CART and ANFIS models is also carried out to appraise the prediction capability of these models. The results show that the ANFIS prediction model can track the change in machine conditions and has the potential for using as a tool to machine fault prognosis.

Multi-modal summarization of key events and top players in sports tournament videos

To detect and annotate the key events of live sports videos, we need to tackle the semantic gaps of audio-visual information. Previous work has successfully extracted semantic from the time-stamped web match reports, which are synchronized with the video contents. However, web and social media articles with no time-stamps have not been fully leveraged, despite they are increasingly used to complement the coverage of major sporting tournaments. This paper aims to address this limitation using a novel multimodal summarization framework that is based on sentiment analysis and players' popularity. It uses audiovisual contents, web articles, blogs, and commentators' speech to automatically annotate and visualize the key events and key players in a sports tournament coverage. The experimental results demonstrate that the automatically generated video summaries are aligned with the events identified from the official website match reports.

Model selection and error estimation

We study model selection strategies based on penalized empirical loss minimization. We point out a tight relationship between error estimation and data-based complexity penalization: any good error estimate may be converted into a data-based penalty function and the performance of the estimate is governed by the quality of the error estimate. We consider several penalty functions, involving error estimates on independent test data, empirical VC dimension, empirical VC entropy, and margin-based quantities. We also consider the maximal difference between the error on the first half of the training data and the second half, and the expected maximal discrepancy, a closely related capacity estimate that can be calculated by Monte Carlo integration. Maximal discrepancy penalty functions are appealing for pattern classification problems, since their computation is equivalent to empirical risk minimization over the training data with some labels flipped.

Rademacher and Gaussian complexities: Risk bounds and structural results

We investigate the use of certain data-dependent estimates of the complexity of a function class, called Rademacher and Gaussian complexities. In a decision theoretic setting, we prove general risk bounds in terms of these complexities. We consider function classes that can be expressed as combinations of functions from basis classes and show how the Rademacher and Gaussian complexities of such a function class can be bounded in terms of the complexity of the basis classes. We give examples of the application of these techniques in finding data-dependent risk bounds for decision trees, neural networks and support vector machines.

Exponentiated gradient algorithms for conditional random fields and max-margin Markov Networks

Log-linear and maximum-margin models are two commonly-used methods in supervised machine learning, and are frequently used in structured prediction problems. Efficient learning of parameters in these models is therefore an important problem, and becomes a key factor when learning from very large data sets. This paper describes exponentiated gradient (EG) algorithms for training such models, where EG updates are applied to the convex dual of either the log-linear or max-margin objective function; the dual in both the log-linear and max-margin cases corresponds to minimizing a convex function with simplex constraints. We study both batch and online variants of the algorithm, and provide rates of convergence for both cases. In the max-margin case, O(1/ε) EG updates are required to reach a given accuracy ε in the dual; in contrast, for log-linear models only O(log(1/ε)) updates are required. For both the max-margin and log-linear cases, our bounds suggest that the online EG algorithm requires a factor of n less computation to reach a desired accuracy than the batch EG algorithm, where n is the number of training examples. Our experiments confirm that the online algorithms are much faster than the batch algorithms in practice. We describe how the EG updates factor in a convenient way for structured prediction problems, allowing the algorithms to be efficiently applied to problems such as sequence learning or natural language parsing. We perform extensive evaluation of the algorithms, comparing them to L-BFGS and stochastic gradient descent for log-linear models, and to SVM-Struct for max-margin models. The algorithms are applied to a multi-class problem as well as to a more complex large-scale parsing task. In all these settings, the EG algorithms presented here outperform the other methods.

Geometry vs appearance for discriminating between posed and spontaneous emotions

Spontaneous facial expressions differ from posed ones in appearance, timing and accompanying head movements. Still images cannot provide timing or head movement information directly. However, indirectly the distances between key points on a face extracted from a still image using active shape models can capture some movement and pose changes. This information is superposed on information about non-rigid facial movement that is also part of the expression. Does geometric information improve the discrimination between spontaneous and posed facial expressions arising from discrete emotions? We investigate the performance of a machine vision system for discrimination between posed and spontaneous versions of six basic emotions that uses SIFT appearance based features and FAP geometric features. Experimental results on the NVIE database demonstrate that fusion of geometric information leads only to marginal improvement over appearance features. Using fusion features, surprise is the easiest emotion (83.4% accuracy) to be distinguished, while disgust is the most difficult (76.1%). Our results find different important facial regions between discriminating posed versus spontaneous version of one emotion and classifying the same emotion versus other emotions. The distribution of the selected SIFT features shows that mouth is more important for sadness, while nose is more important for surprise, however, both the nose and mouth are important for disgust, fear, and happiness. Eyebrows, eyes, nose and mouth are important for anger.

Evaluation of texture and geometry for dimensional facial expression recognition

Facial expression recognition (FER) algorithms mainly focus on classification into a small discrete set of emotions or representation of emotions using facial action units (AUs). Dimensional representation of emotions as continuous values in an arousal-valence space is relatively less investigated. It is not fully known whether fusion of geometric and texture features will result in better dimensional representation of spontaneous emotions. Moreover, the performance of many previously proposed approaches to dimensional representation has not been evaluated thoroughly on publicly available databases. To address these limitations, this paper presents an evaluation framework for dimensional representation of spontaneous facial expressions using texture and geometric features. SIFT, Gabor and LBP features are extracted around facial fiducial points and fused with FAP distance features. The CFS algorithm is adopted for discriminative texture feature selection. Experimental results evaluated on the publicly accessible NVIE database demonstrate that fusion of texture and geometry does not lead to a much better performance than using texture alone, but does result in a significant performance improvement over geometry alone. LBP features perform the best when fused with geometric features. Distributions of arousal and valence for different emotions obtained via the feature extraction process are compared with those obtained from subjective ground truth values assigned by viewers. Predicted valence is found to have a more similar distribution to ground truth than arousal in terms of covariance or Bhattacharya distance, but it shows a greater distance between the means.