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.

Data-driven impostor selection for T-norm score normalisation and the background dataset in SVM-based speaker verification

A data-driven background dataset refinement technique was recently proposed for SVM based speaker verification. This method selects a refined SVM background dataset from a set of candidate impostor examples after individually ranking examples by their relevance. This paper extends this technique to the refinement of the T-norm dataset for SVM-based speaker verification. The independent refinement of the background and T-norm datasets provides a means of investigating the sensitivity of SVM-based speaker verification performance to the selection of each of these datasets. Using refined datasets provided improvements of 13% in min. DCF and 9% in EER over the full set of impostor examples on the 2006 SRE corpus with the majority of these gains due to refinement of the T-norm dataset. Similar trends were observed for the unseen data of the NIST 2008 SRE.

Scatter difference NAP for SVM speaker recognition

This paper presents Scatter Difference Nuisance Attribute Projection (SD-NAP) as an enhancement to NAP for SVM-based speaker verification. While standard NAP may inadvertently remove desirable speaker variability, SD-NAP explicitly de-emphasises this variability by incorporating a weighted version of the between-class scatter into the NAP optimisation criterion. Experimental evaluation of SD-NAP with a variety of SVM systems on the 2006 and 2008 NIST SRE corpora demonstrate that SD-NAP provides improved verification performance over standard NAP in most cases, particularly at the EER operating point.

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.

Data-driven background dataset selection for SVM-based speaker verification

The recently proposed data-driven background dataset refinement technique provides a means of selecting an informative background for support vector machine (SVM)-based speaker verification systems. This paper investigates the characteristics of the impostor examples in such highly-informative background datasets. Data-driven dataset refinement individually evaluates the suitability of candidate impostor examples for the SVM background prior to selecting the highest-ranking examples as a refined background dataset. Further, the characteristics of the refined dataset were analysed to investigate the desired traits of an informative SVM background. The most informative examples of the refined dataset were found to consist of large amounts of active speech and distinctive language characteristics. The data-driven refinement technique was shown to filter the set of candidate impostor examples to produce a more disperse representation of the impostor population in the SVM kernel space, thereby reducing the number of redundant and less-informative examples in the background dataset. Furthermore, data-driven refinement was shown to provide performance gains when applied to the difficult task of refining a small candidate dataset that was mis-matched to the evaluation conditions.

Machine prognostics based on health state estimation using SVM

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 safety. In condition-based maintenance (CBM), maintenance is performed based on information collected through condition monitoring and an 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. All machine components are subjected to degradation processes in real environments and they have certain failure characteristics which can be related to the operating conditions. This paper describes a technique for accurate assessment of the remnant life of machines based on health state probability estimation and involving historical knowledge embedded in the closed loop diagnostics and prognostics systems. The technique uses a Support Vector Machine (SVM) classifier as a tool for estimating health state probability of machine degradation, which can affect the accuracy of prediction. To validate the feasibility of the proposed model, real life historical data from bearings of High Pressure Liquefied Natural Gas (HP-LNG) pumps were analysed and used to obtain the optimal prediction of remaining useful life. The results obtained were very encouraging and showed that the proposed prognostic system based on health state probability estimation has the potential to be used as an estimation tool for remnant life prediction in industrial machinery.

A two-stage SVM method to predict membrane protein types by incorporating amino acid classifications and physicochemical properties into a general form of Chou's PseAAC

Membrane proteins play important roles in many biochemical processes and are also attractive targets of drug discovery for various diseases. The elucidation of membrane protein types provides clues for understanding the structure and function of proteins. Recently we developed a novel system for predicting protein subnuclear localizations. In this paper, we propose a simplified version of our system for predicting membrane protein types directly from primary protein structures, which incorporates amino acid classifications and physicochemical properties into a general form of pseudo-amino acid composition. In this simplified system, we will design a two-stage multi-class support vector machine combined with a two-step optimal feature selection process, which proves very effective in our experiments. The performance of the present method is evaluated on two benchmark datasets consisting of five types of membrane proteins. The overall accuracies of prediction for five types are 93.25% and 96.61% via the jackknife test and independent dataset test, respectively. These results indicate that our method is effective and valuable for predicting membrane protein types. A web server for the proposed method is available at http://www.juemengt.com/jcc/memty_page.php

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.

Prognosis of bearing failure based on health state estimation

This paper proposes a new prognosis model based on the technique for health state estimation of machines for accurate assessment of the remnant life. For the evaluation of health stages of machines, the Support Vector Machine (SVM) classifier was employed to obtain the probability of each health state. Two case studies involving bearing failures were used to validate the proposed model. Simulated bearing failure data and experimental data from an accelerated bearing test rig were used to train and test the model. The result obtained is very encouraging and shows that the proposed prognostic model produces promising results and has the potential to be used as an estimation tool for machine remnant life prediction.

Integrated Diagnosis and Prognosis Model for High Pressure LNG Pump

In condition-based maintenance (CBM), effective diagnostics and prognostics are essential tools for maintenance engineers to identify imminent fault and to predict the remaining useful life before the components finally fail. This enables remedial actions to be taken in advance and reschedules production if necessary. This paper presents a technique for accurate assessment of the remnant life of machines based on historical failure knowledge embedded in the closed loop diagnostic and prognostic system. The technique uses the Support Vector Machine (SVM) classifier for both fault diagnosis and evaluation of health stages of machine degradation. To validate the feasibility of the proposed model, the five different level data of typical four faults from High Pressure Liquefied Natural Gas (HP-LNG) pumps were used for multi-class fault diagnosis. In addition, two sets of impeller-rub data were analysed and employed to predict the remnant life of pump based on estimation of health state. The results obtained were very encouraging and showed that the proposed prognosis system has the potential to be used as an estimation tool for machine remnant life prediction in real life industrial applications.

Rough set-based reasoning and pattern mining for information filtering

An information filtering (IF) system monitors an incoming document stream to find the documents that match the information needs specified by the user profiles. To learn to use the user profiles effectively is one of the most challenging tasks when developing an IF system. With the document selection criteria better defined based on the users’ needs, filtering large streams of information can be more efficient and effective. To learn the user profiles, term-based approaches have been widely used in the IF community because of their simplicity and directness. Term-based approaches are relatively well established. However, these approaches have problems when dealing with polysemy and synonymy, which often lead to an information overload problem. Recently, pattern-based approaches (or Pattern Taxonomy Models (PTM) [160]) have been proposed for IF by the data mining community. These approaches are better at capturing sematic information and have shown encouraging results for improving the effectiveness of the IF system. On the other hand, pattern discovery from large data streams is not computationally efficient. Also, these approaches had to deal with low frequency pattern issues. The measures used by the data mining technique (for example, “support” and “confidences”) to learn the profile have turned out to be not suitable for filtering. They can lead to a mismatch problem. This thesis uses the rough set-based reasoning (term-based) and pattern mining approach as a unified framework for information filtering to overcome the aforementioned problems. This system consists of two stages - topic filtering and pattern mining stages. The topic filtering stage is intended to minimize information overloading by filtering out the most likely irrelevant information based on the user profiles. A novel user-profiles learning method and a theoretical model of the threshold setting have been developed by using rough set decision theory. The second stage (pattern mining) aims at solving the problem of the information mismatch. This stage is precision-oriented. A new document-ranking function has been derived by exploiting the patterns in the pattern taxonomy. The most likely relevant documents were assigned higher scores by the ranking function. Because there is a relatively small amount of documents left after the first stage, the computational cost is markedly reduced; at the same time, pattern discoveries yield more accurate results. The overall performance of the system was improved significantly. The new two-stage information filtering model has been evaluated by extensive experiments. Tests were based on the well-known IR bench-marking processes, using the latest version of the Reuters dataset, namely, the Reuters Corpus Volume 1 (RCV1). The performance of the new two-stage model was compared with both the term-based and data mining-based IF models. The results demonstrate that the proposed information filtering system outperforms significantly the other IF systems, such as the traditional Rocchio IF model, the state-of-the-art term-based models, including the BM25, Support Vector Machines (SVM), and Pattern Taxonomy Model (PTM).

QUT speaker identity verification system for EVALITA 2009

This document outlines the system submitted by the Speech and Audio Research Laboratory at the Queensland University of Technology (QUT) for the Speaker Identity Verication: Application task of EVALITA 2009. This submission consisted of a score-level fusion of three component systems, a joint-factor GMM system and two SVM systems using GLDS and GMM supervector kernels. Development and evaluation results are presented, demonstrating the effectiveness of this fused system approach.

Exploiting multiple feature sets in data-driven impostor dataset selection for speaker verification

This study assesses the recently proposed data-driven background dataset refinement technique for speaker verification using alternate SVM feature sets to the GMM supervector features for which it was originally designed. The performance improvements brought about in each trialled SVM configuration demonstrate the versatility of background dataset refinement. This work also extends on the originally proposed technique to exploit support vector coefficients as an impostor suitability metric in the data-driven selection process. Using support vector coefficients improved the performance of the refined datasets in the evaluation of unseen data. Further, attempts are made to exploit the differences in impostor example suitability measures from varying features spaces to provide added robustness.

Analysis and detection of cognitive load and frustration in drivers' speech

Non-driving related cognitive load and variations of emotional state may impact a driver’s capability to control a vehicle and introduces driving errors. Availability of reliable cognitive load and emotion detection in drivers would benefit the design of active safety systems and other intelligent in-vehicle interfaces. In this study, speech produced by 68 subjects while driving in urban areas is analyzed. A particular focus is on speech production differences in two secondary cognitive tasks, interactions with a co-driver and calls to automated spoken dialog systems (SDS), and two emotional states during the SDS interactions - neutral/negative. A number of speech parameters are found to vary across the cognitive/emotion classes. Suitability of selected cepstral- and production-based features for automatic cognitive task/emotion classification is investigated. A fusion of GMM/SVM classifiers yields an accuracy of 94.3% in cognitive task and 81.3% in emotion classification.

Multilevel converter structure and control

In recent years, multilevel converters are becoming more popular and attractive than traditional converters in high voltage and high power applications. Multilevel converters are particularly suitable for harmonic reduction in high power applications where semiconductor devices are not able to operate at high switching frequencies or in high voltage applications where multilevel converters reduce the need to connect devices in series to achieve high switch voltage ratings. This thesis investigated two aspects of multilevel converters: structure and control. The first part of this thesis focuses on inductance between a DC supply and inverter components in order to minimise loop inductance, which causes overvoltages and stored energy losses during switching. Three dimensional finite element simulations and experimental tests have been carried out for all sections to verify theoretical developments. The major contributions of this section of the thesis are as follows: The use of a large area thin conductor sheet with a rectangular cross section separated by dielectric sheets (planar busbar) instead of circular cross section wires, contributes to a reduction of the stray inductance. A number of approximate equations exist for calculating the inductance of a rectangular conductor but an assumption was made that the current density was uniform throughout the conductors. This assumption is not valid for an inverter with a point injection of current. A mathematical analysis of a planar bus bar has been performed at low and high frequencies and the inductance and the resistance values between the two points of the planar busbar have been determined. A new physical structure for a voltage source inverter with symmetrical planar bus bar structure called Reduced Layer Planar Bus bar, is proposed in this thesis based on the current point injection theory. This new type of planar busbar minimises the variation in stray inductance for different switching states. The reduced layer planar busbar is a new innovation in planar busbars for high power inverters with minimum separation between busbars, optimum stray inductance and improved thermal performances. This type of the planar busbar is suitable for high power inverters, where the voltage source is supported by several capacitors in parallel in order to provide a low ripple DC voltage during operation. A two layer planar busbar with different materials has been analysed theoretically in order to determine the resistance of bus bars during switching. Increasing the resistance of the planar busbar can gain a damping ratio between stray inductance and capacitance and affects the performance of current loop during switching. The aim of this section is to increase the resistance of the planar bus bar at high frequencies (during switching) and without significantly increasing the planar busbar resistance at low frequency (50 Hz) using the skin effect. This contribution shows a novel structure of busbar suitable for high power applications where high resistance is required at switching times. In multilevel converters there are different loop inductances between busbars and power switches associated with different switching states. The aim of this research is to consider all combinations of the switching states for each multilevel converter topology and identify the loop inductance for each switching state. Results show that the physical layout of the busbars is very important for minimisation of the loop inductance at each switch state. Novel symmetrical busbar structures are proposed for multilevel converters with diode-clamp and flying-capacitor topologies which minimise the worst case in stray inductance for different switching states. Overshoot voltages and thermal problems are considered for each topology to optimise the planar busbar structure. In the second part of the thesis, closed loop current techniques have been investigated for single and three phase multilevel converters. The aims of this section are to investigate and propose suitable current controllers such as hysteresis and predictive techniques for multilevel converters with low harmonic distortion and switching losses. This section of the thesis can be classified into three parts as follows: An optimum space vector modulation technique for a three-phase voltage source inverter based on a minimum-loss strategy is proposed. One of the degrees of freedom for optimisation of the space vector modulation is the selection of the zero vectors in the switching sequence. This new method improves switching transitions per cycle for a given level of distortion as the zero vector does not alternate between each sector. The harmonic spectrum and weighted total harmonic distortion for these strategies are compared and results show up to 7% weighted total harmonic distortion improvement over the previous minimum-loss strategy. The concept of SVM technique is a very convenient representation of a set of three-phase voltages or currents used for current control techniques. A new hysteresis current control technique for a single-phase multilevel converter with flying-capacitor topology is developed. This technique is based on magnitude and time errors to optimise the level change of converter output voltage. This method also considers how to improve unbalanced voltages of capacitors using voltage vectors in order to minimise switching losses. Logic controls require handling a large number of switches and a Programmable Logic Device (PLD) is a natural implementation for state transition description. The simulation and experimental results describe and verify the current control technique for the converter. A novel predictive current control technique is proposed for a three-phase multilevel converter, which controls the capacitors' voltage and load current with minimum current ripple and switching losses. The advantage of this contribution is that the technique can be applied to more voltage levels without significantly changing the control circuit. The three-phase five-level inverter with a pure inductive load has been implemented to track three-phase reference currents using analogue circuits and a programmable logic device.