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.

V1-inspired features induce a weighted margin in SVMs

Image representations derived from simplified models of the primary visual cortex (V1), such as HOG and SIFT, elicit good performance in a myriad of visual classification tasks including object recognition/detection, pedestrian detection and facial expression classification. A central question in the vision, learning and neuroscience communities regards why these architectures perform so well. In this paper, we offer a unique perspective to this question by subsuming the role of V1-inspired features directly within a linear support vector machine (SVM). We demonstrate that a specific class of such features in conjunction with a linear SVM can be reinterpreted as inducing a weighted margin on the Kronecker basis expansion of an image. This new viewpoint on the role of V1-inspired features allows us to answer fundamental questions on the uniqueness and redundancies of these features, and offer substantial improvements in terms of computational and storage efficiency.

Improved GMM-based speaker verification using SVM-driven impostor dataset selection

The problem of impostor dataset selection for GMM-based speaker verification is addressed through the recently proposed data-driven background dataset refinement technique. The SVM-based refinement technique selects from a candidate impostor dataset those examples that are most frequently selected as support vectors when training a set of SVMs on a development corpus. This study demonstrates the versatility of dataset refinement in the task of selecting suitable impostor datasets for use in GMM-based speaker verification. The use of refined Z- and T-norm datasets provided performance gains of 15% in EER in the NIST 2006 SRE over the use of heuristically selected datasets. The refined datasets were shown to generalise well to the unseen data of the NIST 2008 SRE.

Improving automatic speaker verification using SVM techniques

Automatic recognition of people is an active field of research with important forensic and security applications. In these applications, it is not always possible for the subject to be in close proximity to the system. Voice represents a human behavioural trait which can be used to recognise people in such situations. Automatic Speaker Verification (ASV) is the process of verifying a persons identity through the analysis of their speech and enables recognition of a subject at a distance over a telephone channel { wired or wireless. A significant amount of research has focussed on the application of Gaussian mixture model (GMM) techniques to speaker verification systems providing state-of-the-art performance. GMM's are a type of generative classifier trained to model the probability distribution of the features used to represent a speaker. Recently introduced to the field of ASV research is the support vector machine (SVM). An SVM is a discriminative classifier requiring examples from both positive and negative classes to train a speaker model. The SVM is based on margin maximisation whereby a hyperplane attempts to separate classes in a high dimensional space. SVMs applied to the task of speaker verification have shown high potential, particularly when used to complement current GMM-based techniques in hybrid systems. This work aims to improve the performance of ASV systems using novel and innovative SVM-based techniques. Research was divided into three main themes: session variability compensation for SVMs; unsupervised model adaptation; and impostor dataset selection. The first theme investigated the differences between the GMM and SVM domains for the modelling of session variability | an aspect crucial for robust speaker verification. Techniques developed to improve the robustness of GMMbased classification were shown to bring about similar benefits to discriminative SVM classification through their integration in the hybrid GMM mean supervector SVM classifier. Further, the domains for the modelling of session variation were contrasted to find a number of common factors, however, the SVM-domain consistently provided marginally better session variation compensation. Minimal complementary information was found between the techniques due to the similarities in how they achieved their objectives. The second theme saw the proposal of a novel model for the purpose of session variation compensation in ASV systems. Continuous progressive model adaptation attempts to improve speaker models by retraining them after exploiting all encountered test utterances during normal use of the system. The introduction of the weight-based factor analysis model provided significant performance improvements of over 60% in an unsupervised scenario. SVM-based classification was then integrated into the progressive system providing further benefits in performance over the GMM counterpart. Analysis demonstrated that SVMs also hold several beneficial characteristics to the task of unsupervised model adaptation prompting further research in the area. In pursuing the final theme, an innovative background dataset selection technique was developed. This technique selects the most appropriate subset of examples from a large and diverse set of candidate impostor observations for use as the SVM background by exploiting the SVM training process. This selection was performed on a per-observation basis so as to overcome the shortcoming of the traditional heuristic-based approach to dataset selection. Results demonstrate the approach to provide performance improvements over both the use of the complete candidate dataset and the best heuristically-selected dataset whilst being only a fraction of the size. The refined dataset was also shown to generalise well to unseen corpora and be highly applicable to the selection of impostor cohorts required in alternate techniques for speaker verification.

Automatically detecting pain in video through facial action units

In a clinical setting, pain is reported either through patient self-report or via an observer. Such measures are problematic as they are: 1) subjective, and 2) give no specific timing information. Coding pain as a series of facial action units (AUs) can avoid these issues as it can be used to gain an objective measure of pain on a frame-by-frame basis. Using video data from patients with shoulder injuries, in this paper, we describe an active appearance model (AAM)-based system that can automatically detect the frames in video in which a patient is in pain. This pain data set highlights the many challenges associated with spontaneous emotion detection, particularly that of expression and head movement due to the patient's reaction to pain. In this paper, we show that the AAM can deal with these movements and can achieve significant improvements in both the AU and pain detection performance compared to the current-state-of-the-art approaches which utilize similarity-normalized appearance features only.

Machine prognostics based on health state probability estimation

The ability to accurately predict the remaining useful life of machine components is critical for machine continuous operation and can also improve productivity and enhance system’s safety. In condition-based maintenance (CBM), maintenance is performed based on information collected through condition monitoring and assessment of the machine health. Effective diagnostics and prognostics are important aspects of CBM for maintenance engineers to schedule a repair and to acquire replacement components before the components actually fail. Although a variety of prognostic methodologies have been reported recently, their application in industry is still relatively new and mostly focused on the prediction of specific component degradations. Furthermore, they required significant and sufficient number of fault indicators to accurately prognose the component faults. Hence, sufficient usage of health indicators in prognostics for the effective interpretation of machine degradation process is still required. Major challenges for accurate longterm prediction of remaining useful life (RUL) still remain to be addressed. Therefore, continuous development and improvement of a machine health management system and accurate long-term prediction of machine remnant life is required in real industry application. This thesis presents an integrated diagnostics and prognostics framework based on health state probability estimation for accurate and long-term prediction of machine remnant life. In the proposed model, prior empirical (historical) knowledge is embedded in the integrated diagnostics and prognostics system for classification of impending faults in machine system and accurate probability estimation of discrete degradation stages (health states). The methodology assumes that machine degradation consists of a series of degraded states (health states) which effectively represent the dynamic and stochastic process of machine failure. The estimation of discrete health state probability for the prediction of machine remnant life is performed using the ability of classification algorithms. To employ the appropriate classifier for health state probability estimation in the proposed model, comparative intelligent diagnostic tests were conducted using five different classifiers applied to the progressive fault data of three different faults in a high pressure liquefied natural gas (HP-LNG) pump. As a result of this comparison study, SVMs were employed in heath state probability estimation for the prediction of machine failure in this research. The proposed prognostic methodology has been successfully tested and validated using a number of case studies from simulation tests to real industry applications. The results from two actual failure case studies using simulations and experiments indicate that accurate estimation of health states is achievable and the proposed method provides accurate long-term prediction of machine remnant life. In addition, the results of experimental tests show that the proposed model has the capability of providing early warning of abnormal machine operating conditions by identifying the transitional states of machine fault conditions. Finally, the proposed prognostic model is validated through two industrial case studies. The optimal number of health states which can minimise the model training error without significant decrease of prediction accuracy was also examined through several health states of bearing failure. The results were very encouraging and show that the proposed prognostic model based on health state probability estimation has the potential to be used as a generic and scalable asset health estimation tool in industrial machinery.

Classification with a reject option using a hinge loss

We consider the problem of binary classification where the classifier can, for a particular cost, choose not to classify an observation. Just as in the conventional classification problem, minimization of the sample average of the cost is a difficult optimization problem. As an alternative, we propose the optimization of a certain convex loss function φ, analogous to the hinge loss used in support vector machines (SVMs). Its convexity ensures that the sample average of this surrogate loss can be efficiently minimized. We study its statistical properties. We show that minimizing the expected surrogate loss—the φ-risk—also minimizes the risk. We also study the rate at which the φ-risk approaches its minimum value. We show that fast rates are possible when the conditional probability P(Y=1|X) is unlikely to be close to certain critical values.

Sparseness vs estimating conditional probabilities : some asymptotic results

One of the nice properties of kernel classifiers such as SVMs is that they often produce sparse solutions. However, the decision functions of these classifiers cannot always be used to estimate the conditional probability of the class label. We investigate the relationship between these two properties and show that these are intimately related: sparseness does not occur when the conditional probabilities can be unambiguously estimated. We consider a family of convex loss functions and derive sharp asymptotic results for the fraction of data that becomes support vectors. This enables us to characterize the exact trade-off between sparseness and the ability to estimate conditional probabilities for these loss functions.

Improved facial expression recognition via uni-hyperplane classification

Large margin learning approaches, such as support vector machines (SVM), have been successfully applied to numerous classification tasks, especially for automatic facial expression recognition. The risk of such approaches however, is their sensitivity to large margin losses due to the influence from noisy training examples and outliers which is a common problem in the area of affective computing (i.e., manual coding at the frame level is tedious so coarse labels are normally assigned). In this paper, we leverage the relaxation of the parallel-hyperplanes constraint and propose the use of modified correlation filters (MCF). The MCF is similar in spirit to SVMs and correlation filters, but with the key difference of optimizing only a single hyperplane. We demonstrate the superiority of MCF over current techniques on a battery of experiments.

Semantic labelling for document feature patterns using ontological subjects

Finding and labelling semantic features patterns of documents in a large, spatial corpus is a challenging problem. Text documents have characteristics that make semantic labelling difficult; the rapidly increasing volume of online documents makes a bottleneck in finding meaningful textual patterns. Aiming to deal with these issues, we propose an unsupervised documnent labelling approach based on semantic content and feature patterns. A world ontology with extensive topic coverage is exploited to supply controlled, structured subjects for labelling. An algorithm is also introduced to reduce dimensionality based on the study of ontological structure. The proposed approach was promisingly evaluated by compared with typical machine learning methods including SVMs, Rocchio, and kNN.

Classification of echolocation calls from 14 species of bat by Support Vector Machines and Ensembles of Neural Networks

Calls from 14 species of bat were classified to genus and species using discriminant function analysis (DFA), support vector machines (SVM) and ensembles of neural networks (ENN). Both SVMs and ENNs outperformed DFA for every species while ENNs (mean identification rate – 97%) consistently outperformed SVMs (mean identification rate – 87%). Correct classification rates produced by the ENNs varied from 91% to 100%; calls from six species were correctly identified with 100% accuracy. Calls from the five species of Myotis, a genus whose species are considered difficult to distinguish acoustically, had correct identification rates that varied from 91 – 100%. Five parameters were most important for classifying calls correctly while seven others contributed little to classification performance.

Acoustic features for multi-level classification of Australian frogs

Over past few decades, frog species have been experiencing dramatic decline around the world. The reason for this decline includes habitat loss, invasive species, climate change and so on. To better know the status of frog species, classifying frogs has become increasingly important. In this study, acoustic features are investigated for multi-level classification of Australian frogs: family, genus and species, including three families, eleven genera and eighty five species which are collected from Queensland, Australia. For each frog species, six instances are selected from which ten acoustic features are calculated. Then, the multicollinearity between ten features are studied for selecting non-correlated features for subsequent analysis. A decision tree (DT) classifier is used to visually and explicitly determine which acoustic features are relatively important for classifying family, which for genus, and which for species. Finally, a weighted support vector machines (SVMs) classifier is used for the multi- level classification with three most important acoustic features respectively. Our experiment results indicate that using different acoustic feature sets can successfully classify frogs at different levels and the average classification accuracy can be up to 85.6%, 86.1% and 56.2% for family, genus and species respectively.

Activity recognition using binary tree SVM

This paper presents an effective classification method based on Support Vector Machines (SVM) in the context of activity recognition. Local features that capture both spatial and temporal information in activity videos have made significant progress recently. Efficient and effective features, feature representation and classification plays a crucial role in activity recognition. For classification, SVMs are popularly used because of their simplicity and efficiency; however the common multi-class SVM approaches applied suffer from limitations including having easily confused classes and been computationally inefficient. We propose using a binary tree SVM to address the shortcomings of multi-class SVMs in activity recognition. We proposed constructing a binary tree using Gaussian Mixture Models (GMM), where activities are repeatedly allocated to subnodes until every new created node contains only one activity. Then, for each internal node a separate SVM is learned to classify activities, which significantly reduces the training time and increases the speed of testing compared to popular the `one-against-the-rest' multi-class SVM classifier. Experiments carried out on the challenging and complex Hollywood dataset demonstrates comparable performance over the baseline bag-of-features method.