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.

Tax morale affects tax compliance : evidence from surveys and an artefactual field experiment

Our working hypotheses is that cross-cultural differences in tax compliance behaviour have foundations in the institutions of tax administration and citizen assessment of the quality of governance. Tax compliance being a complex behavioural issue. Its investigation requires use of a variety of methods and data sources. Results from artefactual field experiments conducted in countries with substantially different political histories and records of governance quality demondtrate that observed differences in tax compliance levels persist over alternative levels of enforcement. The experimental results are shown to be robust by replicating them for the same countries using survey response measures of tax compliance.

Inequality Aversion and Performance in and on the Field

The experimental literature and studies using survey data have established that people care a great deal about their relative economic position and not solely, as standard economic theory assumes, about their absolute economic position. Individuals are concerned about social comparisons. However, behavioral evidence in the field is rare. This paper provides an empirical analysis, testing the model of inequality aversion using two unique panel data sets for basketball and soccer players. We find support that the concept of inequality aversion helps to understand how the relative income situation affects performance in a real competitive environment with real tasks and real incentives.

Temperature and electric field dependent mobility in poly„3-hexylthiophene diodes

Current-voltage (I-V) curves of Poly(3-hexyl-thiophene) (P3HT) diodes have been collected to investigate the polymer hole-dominated charge transport. At room temperature and at low electric fields the I-V characteristic is purely Ohmic whereas at medium-high electric fields, experimental data shows that the hole transport is Trap Dominated - Space Charge Limited Current (TD-SCLC). In this regime, it is possible to extract the I-V characteristic of the P3HT/Al junction showing the ideal Schottky diode behaviour over five orders of magnitude. At high-applied electric fields, holes’ transport is found to be in the trap free SCLC regime. We have measured and modelled in this regime the holes’ mobility to evaluate its dependence from the electric field applied and the temperature of the device.

Temporal-based support vector machines for facial expression recognition

When classifying a signal, ideally we want our classifier to trigger a large response when it encounters a positive example and have little to no response for all other examples. Unfortunately in practice this does not occur with responses fluctuating, often causing false alarms. There exists a myriad of reasons why this is the case, most notably not incorporating the dynamics of the signal into the classification. In facial expression recognition, this has been highlighted as one major research question. In this paper we present a novel technique which incorporates the dynamics of the signal which can produce a strong response when the peak expression is found and essentially suppresses all other responses as much as possible. We conducted preliminary experiments on the extended Cohn-Kanade (CK+) database which shows its benefits. The ability to automatically and accurately recognize facial expressions of drivers is highly relevant to the automobile. For example, the early recognition of “surprise” could indicate that an accident is about to occur; and various safeguards could immediately be deployed to avoid or minimize injury and damage. In this paper, we conducted initial experiments on the extended Cohn-Kanade (CK+) database which shows its benefits.

Field and Service Robotics Results of the 5th International Conference

The 5th International Conference on Field and Service Robotics (FSR05) was held in Port Douglas, Australia, on 29th - 31st July 2005, and brought together the worlds' leading experts in field and service automation. The goal of the conference was to report and encourage the latest research and practical results towards the use of field and service robotics in the community with particular focus on proven technology. The conference provided a forum for researchers, professionals and robot manufacturers to exchange up-to-date technical knowledge and experience. Field robots are robots which operate in outdoor, complex, and dynamic environments. Service robots are those that work closely with humans, with particular applications involving indoor and structured environments. There are a wide range of topics presented in this issue on field and service robots including: Agricultural and Forestry Robotics, Mining and Exploration Robots, Robots for Construction, Security & Defence Robots, Cleaning Robots, Autonomous Underwater Vehicles and Autonomous Flying Robots. This meeting was the fifth in the series and brings FSR back to Australia where it was first held. FSR has been held every 2 years, starting with Canberra 1997, followed by Pittsburgh 1999, Helsinki 2001 and Lake Yamanaka 2003.

In the Rough : In-Field evaluation of an autonomous vehicle for golf course maintenance

Maintenance is a time consuming and expensive task for any golf course or driving range manager. For a golf course the primary tasks are grass mowing and maintenance (fertilizer and herbicide spreading), while for a driving range mowing, maintenance and ball collection are required. All these tasks require an operator to drive a vehicle along paths which are generally predefined. This paper presents some preliminary in-field tsting results for an automated tractor vehicle performing golf ball collection on an actual driving range, and mowing on difficult unstructured terrain.

A point interpolation method with locally smoothed strain field (PIM-LS2) for mechanics problems using triangular mesh

A point interpolation method with locally smoothed strain field (PIM-LS2) is developed for mechanics problems using a triangular background mesh. In the PIM-LS2, the strain within each sub-cell of a nodal domain is assumed to be the average strain over the adjacent sub-cells of the neighboring element sharing the same field node. We prove theoretically that the energy norm of the smoothed strain field in PIM-LS2 is equivalent to that of the compatible strain field, and then prove that the solution of the PIM- LS2 converges to the exact solution of the original strong form. Furthermore, the softening effects of PIM-LS2 to system and the effects of the number of sub-cells that participated in the smoothing operation on the convergence of PIM-LS2 are investigated. Intensive numerical studies verify the convergence, softening effects and bound properties of the PIM-LS2, and show that the very ‘‘tight’’ lower and upper bound solutions can be obtained using PIM-LS2.

Avoiding full extension field arithmetic in pairing computations

The most costly operations encountered in pairing computations are those that take place in the full extension field Fpk . At high levels of security, the complexity of operations in Fpk dominates the complexity of the operations that occur in the lower degree subfields. Consequently, full extension field operations have the greatest effect on the runtime of Miller’s algorithm. Many recent optimizations in the literature have focussed on improving the overall operation count by presenting new explicit formulas that reduce the number of subfield operations encountered throughout an iteration of Miller’s algorithm. Unfortunately, almost all of these improvements tend to suffer for larger embedding degrees where the expensive extension field operations far outweigh the operations in the smaller subfields. In this paper, we propose a new way of carrying out Miller’s algorithm that involves new explicit formulas which reduce the number of full extension field operations that occur in an iteration of the Miller loop, resulting in significant speed ups in most practical situations of between 5 and 30 percent.

Delaying mismatched field multiplications in pairing computations

Miller’s algorithm for computing pairings involves perform- ing multiplications between elements that belong to different finite fields. Namely, elements in the full extension field Fpk are multiplied by elements contained in proper subfields F pk/d , and by elements in the base field Fp . We show that significant speedups in pairing computations can be achieved by delaying these “mismatched” multiplications for an optimal number of iterations. Importantly, we show that our technique can be easily integrated into traditional pairing algorithms; implementers can exploit the computational savings herein by applying only minor changes to existing pairing code.

Large margin vector quantization

In this paper we describe the Large Margin Vector Quantization algorithm (LMVQ), which uses gradient ascent to maximise the margin of a radial basis function classifier. We present a derivation of the algorithm, which proceeds from an estimate of the class-conditional probability densities. We show that the key behaviour of Kohonen's well-known LVQ2 and LVQ3 algorithms emerge as natural consequences of our formulation. We compare the performance of LMVQ with that of Kohonen's LVQ algorithms on an artificial classification problem and several well known benchmark classification tasks. We find that the classifiers produced by LMVQ attain a level of accuracy that compares well with those obtained via LVQ1, LVQ2 and LVQ3, with reduced storage complexity. We indicate future directions of enquiry based on the large margin approach to Learning Vector Quantization.

Researching crime and violence: Untold stories from the field

Undertaking empirical research on crime and violence can be a tricky enterprise fraught with ethical, methodological, intellectual and legal implications. This chapter takes readers on a reflective journey through the qualitative methodologies I used to research sex work in Kings Cross, miscarriages of justice, female delinquency, sexual violence, and violence in rural and regional settings over a period of nearly 30 years. Reflecting on these experiences, the chapter explores and analyses the reality of doing qualitative field research, the role of the researcher, the politics of subjectivity, the exercise of power, and the ‘muddiness’ of the research process, which is often overlooked in sanitised accounts of the research process (Byrne-Armstrong, Higgs and Horsfall, 2001; Davies, 2000).