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.

Customer-oriented deviance (COD) : a conceptualisation, development and validation of a behavioural measure

Frontline employee behaviours are recognised as vital for achieving a competitive advantage for service organisations. The services marketing literature has comprehensively examined ways to improve frontline employee behaviours in service delivery and recovery. However, limited attention has been paid to frontline employee behaviours that favour customers in ways that go against organisational norms or rules. This study examines these behaviours by introducing a behavioural concept of Customer-Oriented Deviance (COD). COD is defined as, “frontline employees exhibiting extra-role behaviours that they perceive to defy existing expectations or prescribed rules of higher authority through service adaptation, communication and use of resources to benefit customers during interpersonal service encounters.” This thesis develops a COD measure and examines the key determinants of these behaviours from a frontline employee perspective. Existing research on similar behaviours that has originated in the positive deviance and pro-social behaviour domains has limitations and is considered inadequate to examine COD in the services context. The absence of a well-developed body of knowledge on non-conforming service behaviours has implications for both theory and practice. The provision of ‘special favours’ increases customer satisfaction but the over-servicing of customers is also counterproductive for the service delivery and costly for the organisation. Despite these implications of non-conforming service behaviours, there is little understanding about the nature of these behaviours and its key drivers. This research builds on inadequacies in prior research on positive deviance, pro-social and pro-customer literature to develop the theoretical foundation of COD. The concept of positive deviance which has predominantly been used to study organisational behaviours is applied within a services marketing setting. Further, it addresses previous limitations in pro-social and pro-customer behavioural literature that has examined limited forms of behaviours with no clear understanding on the nature of these behaviours. Building upon these literature streams, this research adopts a holistic approach towards the conceptualisation of COD. It addresses previous shortcomings in the literature by providing a well bounded definition, developing a psychometrically sound measure of COD and a conceptually well-founded model of COD. The concept of COD was examined across three separate studies and based on the theoretical foundations of role theory and social identity theory. Study 1 was exploratory and based on in-depth interviews using the Critical Incident Technique (CIT). The aim of Study 1 was to understand the nature of COD and qualitatively identify its key drivers. Thematic analysis was conducted to analyse the data and the two potential dimensions of COD behaviours of Deviant Service Adaptation (DSA) and Deviant Service Communication (DSC) were revealed in the analysis. In addition, themes representing the potential influences of COD were broadly classified as individual factors, situational factors, and organisational factors. Study 2 was a scale development procedure that involved the generation and purification of items for the measure based on two student samples working in customer service roles (Pilot sample, N=278; Initial validation sample, N=231). The results for the reliability and Exploratory Factor Analyses (EFA) on the pilot sample suggested the scale had poor psychometric properties. As a result, major revisions were made in terms of item wordings and new items were developed based on the literature to reflect a new dimension, Deviant Use of Resources (DUR). The revised items were tested on the initial validation sample with the EFA analysis suggesting a four-factor structure of COD. The aim of Study 3 was to further purify the COD measure and test for nomological validity based on its theoretical relationships with key antecedents and similar constructs (key correlates). The theoretical model of COD consisting of nine hypotheses was tested on a retail and hospitality sample of frontline employees (Retail N=311; Hospitality N=305) of a market research panel using an online survey. The data was analysed using Structural Equation Modelling (SEM). The results provided support for a re-specified second-order three-factor model of COD which consists of 11 items. Overall, the COD measure was found to be reliable and valid, demonstrating convergent validity, discriminant validity and marginal partial invariance for the factor loadings. The results showed support for nomological validity, although the antecedents had differing impact on COD across samples. Specifically, empathy and perspective-taking, role conflict, and job autonomy significantly influenced COD in the retail sample, whereas empathy and perspective-taking, risk-taking propensity and role conflict were significant predictors in the hospitality sample. In addition, customer orientation-selling orientation, the altruistic dimension of organisational citizenship behaviours, workplace deviance, and social desirability responding were found to correlate with COD. This research makes several contributions to theory. First, the findings of this thesis extend the literature on positive deviance, pro-social and pro-customer behaviours. Second, the research provides an empirically tested model which describes the antecedents of COD. Third, this research contributes by providing a reliable and valid measure of COD. Finally, the research investigates the differential effects of the key antecedents in different service sectors on COD. The research findings also contribute to services marketing practice. Based on the research findings, service practitioners can better understand the phenomenon of COD and utilise the measurement tool to calibrate COD levels within their organisations. Knowledge on the key determinants of COD will help improve recruitment and training programs and drive internal initiatives within the firm.

Virtual Environment Visualisation of Executable Business Process Models

Business Process Modelling is a fast growing field in business and information technology, which uses visual grammars to model and execute the processes within an organisation. However, many analysts present such models in a 2D static and iconic manner that is difficult to understand by many stakeholders. Difficulties in understanding such grammars can impede the improvement of processes within an enterprise due to communication problems. In this chapter we present a novel framework for intuitively visualising animated business process models in interactive Virtual Environments. We also show that virtual environment visualisations can be performed with present 2D business process modelling technology, thus providing a low barrier to entry for business process practitioners. Two case studies are presented from film production and healthcare domains that illustrate the ease with which these visualisations can be created. This approach can be generalised to other executable workflow systems, for any application domain being modelled.

Behavioural biases and information disclosure laws relating to residential property sales : narrowing the gap between existing laws and calls for future reforms

Market failures involving the sale of complex merchandise, such as residential property, financial products and credit, have principally been attributed to information asymmetries. Existing legislative and regulatory responses were developed having regard to consumer protection policies based on traditional economic theories that focus on the notion of the ‘rational consumer’. Governmental responses therefore seek to impose disclosure obligations on sellers of complex goods or products to ensure that consumers have sufficient information upon which to make a decision. Emergent research, based on behavioural economics, challenges traditional ideas and instead focuses on the actual behaviour of consumers. This approach suggests that consumers as a whole do not necessarily benefit from mandatory disclosure because some, if not most, consumers do not pay attention to the disclosed information before they make a decision to purchase. The need for consumer policies to take consumer characteristics and behaviour into account is being increasingly recognised by governments, and most recently in the policy framework suggested by the Australian Productivity Commission

Driver behaviour and decision making at railway level crossings : An exploratory on-road case study

Crashes at rail level crossings represent a significant problem, both in Australia and worldwide. Advances in driving assessment methods, such as the provision of on-road instrumented test vehicles, now provide researchers with the opportunity to further understand driver behaviour at rail level crossings in ways not previously possible. This paper gives an overview of a recent on-road pilot study of driver behaviour at rail level crossings in which 25 participants drove a pre-determined route, incorporating 4 rail level crossings, using MUARC's instrumented On-Road Test Vehicle (ORTeV). Drivers provided verbal commentary whilst driving the route, and a range of other data were collected, including eye fixations, forward, cockpit and driver video, and vehicle data (speed, braking, steering wheel angle, lane tracking etc). Participants also completed a post trial cognitive task analysis interview. Extracts from the wider analyses are used to examine in depth driver behaviour at one of the rail level crossings encountered during the study. The analysis presented, along with the overall analysis undertaken, gives insight into the driver and wider systems factors that shape behaviour at rail level crossings, and highlights the utility of using a multi-method, instrumented vehicle approach for gathering data regarding driver behaviour in different contexts.

Are organ donation communication decisions reasoned or reactive? A test of the utility of an augmented Theory of Planned Behaviour with the Prototype/Willingness Model

Objectives: To explore whether people's organ donation consent decisions occur via a reasoned and/or social reaction pathway. --------- Design: We examined prospectively students' and community members' decisions to register consent on a donor register and discuss organ donation wishes with family. --------- Method: Participants completed items assessing theory of planned behaviour (TPB; attitude, subjective norm, perceived behavioural control (PBC)), prototype/willingness model (PWM; donor prototype favourability/similarity, past behaviour), and proposed additional influences (moral norm, self-identity, recipient prototypes) for registering (N=339) and discussing (N=315) intentions/willingness. Participants self-reported their registering (N=177) and discussing (N=166) behaviour 1 month later. The utility of the (1) TPB, (2) PWM, (3) augmented TPB with PWM, and (4) augmented TPB with PWM and extensions was tested using structural equation modelling for registering and discussing intentions/willingness, and logistic regression for behaviour. --------- Results: While the TPB proved a more parsimonious model, fit indices suggested that the other proposed models offered viable options, explaining greater variance in communication intentions/willingness. The TPB, augmented TPB with PWM, and extended augmented TPB with PWM best explained registering and discussing decisions. The proposed and revised PWM also proved an adequate fit for discussing decisions. Respondents with stronger intentions (and PBC for registering) had a higher likelihood of registering and discussing. --------- Conclusions: People's decisions to communicate donation wishes may be better explained via a reasoned pathway (especially for registering); however, discussing involves more reactive elements. The role of moral norm, self-identity, and prototypes as influences predicting communication decisions were highlighted also.

Hippocampal models for simultaneous localisation and mapping on an autonomous robot

To navigate successfully in a novel environment a robot needs to be able to Simultaneously Localize And Map (SLAM) its surroundings. The most successful solutions to this problem so far have involved probabilistic algorithms, but there has been much promising work involving systems based on the workings of part of the rodent brain known as the hippocampus. In this paper we present a biologically plausible system called RatSLAM that uses competitive attractor networks to carry out SLAM in a probabilistic manner. The system can effectively perform parameter self-calibration and SLAM in one dimension. Tests in two dimensional environments revealed the inability of the RatSLAM system to maintain multiple pose hypotheses in the face of ambiguous visual input. These results support recent rat experimentation that suggest current competitive attractor models are not a complete solution to the hippocampal modelling problem.

A Helicopter named Dolly : behavioural cloning for autonomous helicopter control

This paper considers the pros and cons of using Behavioural cloning for the development of low-level helicopter automation modules. Over the course of this project several Behavioural cloning approaches have been investigated. The results of the most effective Behavioural cloning approach are then compared to PID modules designed for the same aircraft. The comparison takes into consideration development time, reliability, and control performance. It has been found that Behavioural cloning techniques employing local approximators and a wide state-space coverage during training can produce stabilising control modules in less time than tuning PID controllers. However, performance and reliabity deficits have been found to exist with the Behavioural Cloning, attributable largely to the time variant nature of the dynamics due to the operating environment, and the pilot actions being poor for teaching. The final conclusion drawn here is that tuning PID modules remains superior to behavioural cloning for low-level helicopter automation.

Emotional and behavioural reaction to intrusive thoughts

A self-report measure of the emotional and behavioural reactions to intrusive thoughts was developed. The paper presents data that confirm the stability, reliability and validity of the new 7-item measure. Emotional and behavioural reactions to intrusions emerged as separate factors on the Emotional and Behavioural Reactions to Intrusions Questionnaire (EBRIQ), a finding confirmed by an independent stress study. Test retest reliability over 30-70 days was good. Expected relationships with other constructs were significant. Stronger negative responses to intrusions were associated with lower mindfulness scores and higher ratings of experiential avoidance, thought suppression and intensity and frequency of craving. The EBRIQ will help explore differences in reactions to intrusive thoughts in clinical and non clinical populations, and across different emotional and behavioural states. It will also be useful in assessing the effects of therapeutic approaches such as mindfulness.

Driving performance on monotonous roads

Driving on motorways has largely been reduced to a lane-keeping task with cruise control. Rapidly, drivers are likely to get bored with such a task and take their attention away from the road. This is of concern in terms of road safety – particularly for professional drivers - since inattention has been identified as one of the main contributing factors to road crashes and is estimated to be involved in 20 to 30% of these crashes. Furthermore, drivers are not aware that their vigilance level has decreased and that their driving performance is impaired. Intelligent Transportation System (ITS) intervention can be used as a countermeasure against vigilance decrement. This paper aims to identify a variety of metrics impacted during monotonous driving - ranging from vehicle data to physiological variables - and relate them to two monotonous factors namely the monotony of the road design (straightness) and the monotony of the environment (landscape, signage, traffic). Data are collected in a driving simulator instrumented with an eye tracking system, a heart rate monitor and an electrodermal activity device (N=25 participants). The two monotonous factors are varied (high and low) leading to the use of four different driving scenarios (40 minutes each). We show with Generalised Linear Mixed Models that driver performance decreases faster when the road is monotonous. We also highlight that road monotony impairs a variety of driving performance and vigilance measures, ranging from speed, lateral position of the vehicle to physiological measurements such as heart rate variability, blink frequency and electrodermal activity. This study informs road designers of the importance of having a varied road environment. It also provides a range of metrics that can be used to detect in real-time the impairment of driving performance on monotonous roads. Such knowledge could result in the development of an in-vehicle device warning drivers at early signs of driving performance impairment on monotonous roads.

Comprehensive evaluation of driver gaze pattern using fuzzy rules for driver training

The over representation of novice drivers in crashes is alarming. Research indicates that one in five drivers’ crashes within their first year of driving. Driver training is one of the interventions aimed at decreasing the number of crashes that involve young drivers. Currently, there is a need to develop comprehensive driver evaluation system that benefits from the advances in Driver Assistance Systems. Since driving is dependent on fuzzy inputs from the driver (i.e. approximate distance calculation from the other vehicles, approximate assumption of the other vehicle speed), it is necessary that the evaluation system is based on criteria and rules that handles uncertain and fuzzy characteristics of the drive. This paper presents a system that evaluates the data stream acquired from multiple in-vehicle sensors (acquired from Driver Vehicle Environment-DVE) using fuzzy rules and classifies the driving manoeuvres (i.e. overtake, lane change and turn) as low risk or high risk. The fuzzy rules use parameters such as following distance, frequency of mirror checks, gaze depth and scan area, distance with respect to lanes and excessive acceleration or braking during the manoeuvre to assess risk. The fuzzy rules to estimate risk are designed after analysing the selected driving manoeuvres performed by driver trainers. This paper focuses mainly on the difference in gaze pattern for experienced and novice drivers during the selected manoeuvres. Using this system, trainers of novice drivers would be able to empirically evaluate and give feedback to the novice drivers regarding their driving behaviour.

Bayesian hybrid algorithms and models : implementation and associated issues

This thesis addresses computational challenges arising from Bayesian analysis of complex real-world problems. Many of the models and algorithms designed for such analysis are ‘hybrid’ in nature, in that they are a composition of components for which their individual properties may be easily described but the performance of the model or algorithm as a whole is less well understood. The aim of this research project is to after a better understanding of the performance of hybrid models and algorithms. The goal of this thesis is to analyse the computational aspects of hybrid models and hybrid algorithms in the Bayesian context. The first objective of the research focuses on computational aspects of hybrid models, notably a continuous finite mixture of t-distributions. In the mixture model, an inference of interest is the number of components, as this may relate to both the quality of model fit to data and the computational workload. The analysis of t-mixtures using Markov chain Monte Carlo (MCMC) is described and the model is compared to the Normal case based on the goodness of fit. Through simulation studies, it is demonstrated that the t-mixture model can be more flexible and more parsimonious in terms of number of components, particularly for skewed and heavytailed data. The study also reveals important computational issues associated with the use of t-mixtures, which have not been adequately considered in the literature. The second objective of the research focuses on computational aspects of hybrid algorithms for Bayesian analysis. Two approaches will be considered: a formal comparison of the performance of a range of hybrid algorithms and a theoretical investigation of the performance of one of these algorithms in high dimensions. For the first approach, the delayed rejection algorithm, the pinball sampler, the Metropolis adjusted Langevin algorithm, and the hybrid version of the population Monte Carlo (PMC) algorithm are selected as a set of examples of hybrid algorithms. Statistical literature shows how statistical efficiency is often the only criteria for an efficient algorithm. In this thesis the algorithms are also considered and compared from a more practical perspective. This extends to the study of how individual algorithms contribute to the overall efficiency of hybrid algorithms, and highlights weaknesses that may be introduced by the combination process of these components in a single algorithm. The second approach to considering computational aspects of hybrid algorithms involves an investigation of the performance of the PMC in high dimensions. It is well known that as a model becomes more complex, computation may become increasingly difficult in real time. In particular the importance sampling based algorithms, including the PMC, are known to be unstable in high dimensions. This thesis examines the PMC algorithm in a simplified setting, a single step of the general sampling, and explores a fundamental problem that occurs in applying importance sampling to a high-dimensional problem. The precision of the computed estimate from the simplified setting is measured by the asymptotic variance of the estimate under conditions on the importance function. Additionally, the exponential growth of the asymptotic variance with the dimension is demonstrated and we illustrates that the optimal covariance matrix for the importance function can be estimated in a special case.

Next Generation ‘Filmmaking’: New Markets, New Methods and New Business Models

Digital production and distribution technologies may create new opportunities for filmmaking in Australia. A culture of new approaches to filmmaking is emerging driven by ‘next generation filmmakers’ who are willing to consider new business models: from online web series to short films produced for mobile phones. At the same time cultural representation itself is transforming within an interactive, social media driven environment. Yet there is very little research into next generation filmmaking. The aim of this paper is to scope and discuss three key aspects of next generation filmmaking, namely: digital trends in film distribution and marketing; processes and strategies of ‘next generation’ filmmakers; and case studies of viable next generation business models and filmmaking practices. We conclude with a brief examination of the implications for media and cultural policy which suggests the future possibility of a rapprochement between creative industries discourse and cultural policy.

Incorporating uncertainty in environmental models informed by imagery

In this thesis, the issue of incorporating uncertainty for environmental modelling informed by imagery is explored by considering uncertainty in deterministic modelling, measurement uncertainty and uncertainty in image composition. Incorporating uncertainty in deterministic modelling is extended for use with imagery using the Bayesian melding approach. In the application presented, slope steepness is shown to be the main contributor to total uncertainty in the Revised Universal Soil Loss Equation. A spatial sampling procedure is also proposed to assist in implementing Bayesian melding given the increased data size with models informed by imagery. Measurement error models are another approach to incorporating uncertainty when data is informed by imagery. These models for measurement uncertainty, considered in a Bayesian conditional independence framework, are applied to ecological data generated from imagery. The models are shown to be appropriate and useful in certain situations. Measurement uncertainty is also considered in the context of change detection when two images are not co-registered. An approach for detecting change in two successive images is proposed that is not affected by registration. The procedure uses the Kolmogorov-Smirnov test on homogeneous segments of an image to detect change, with the homogeneous segments determined using a Bayesian mixture model of pixel values. Using the mixture model to segment an image also allows for uncertainty in the composition of an image. This thesis concludes by comparing several different Bayesian image segmentation approaches that allow for uncertainty regarding the allocation of pixels to different ground components. Each segmentation approach is applied to a data set of chlorophyll values and shown to have different benefits and drawbacks depending on the aims of the analysis.