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.

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.

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.

An innovative information gathering and data analysis platform for railway level crossing safety data

This paper describes an innovative platform that facilitates the collection of objective safety data around occurrences at railway level crossings using data sources including forward-facing video, telemetry from trains and geo-referenced asset and survey data. This platform is being developed with support by the Australian rail industry and the Cooperative Research Centre for Rail Innovation. The paper provides a description of the underlying accident causation model, the development methodology and refinement process as well as a description of the data collection platform. The paper concludes with a brief discussion of benefits this project is expected to provide the Australian rail industry.

High body mass index is not a barrier to physical activity: Analysis of international rugby players' anthropometric data

Recent data indicate that levels of overweight and obesity are increasing at an alarming rate throughout the world. At a population level (and commonly to assess individual health risk), the prevalence of overweight and obesity is calculated using cut-offs of the Body Mass Index (BMI) derived from height and weight. Similarly, the BMI is also used to classify individuals and to provide a notional indication of potential health risk. It is likely that epidemiologic surveys that are reliant on BMI as a measure of adiposity will overestimate the number of individuals in the overweight (and slightly obese) categories. This tendency to misclassify individuals may be more pronounced in athletic populations or groups in which the proportion of more active individuals is higher. This differential is most pronounced in sports where it is advantageous to have a high BMI (but not necessarily high fatness). To illustrate this point we calculated the BMIs of international professional rugby players from the four teams involved in the semi-finals of the 2003 Rugby Union World Cup. According to the World Health Organisation (WHO) cut-offs for BMI, approximately 65% of the players were classified as overweight and approximately 25% as obese. These findings demonstrate that a high BMI is commonplace (and a potentially desirable attribute for sport performance) in professional rugby players. An unanswered question is what proportion of the wider population, classified as overweight (or obese) according to the BMI, is misclassified according to both fatness and health risk? It is evident that being overweight should not be an obstacle to a physically active lifestyle. Similarly, a reliance on BMI alone may misclassify a number of individuals who might otherwise have been automatically considered fat and/or unfit.

A Petrov-Galerkin method for a singularly perturbed ordinary differential equation with non-smooth data

In this paper, a singularly perturbed ordinary differential equation with non-smooth data is considered. The numerical method is generated by means of a Petrov-Galerkin finite element method with the piecewise-exponential test function and the piecewise-linear trial function. At the discontinuous point of the coefficient, a special technique is used. The method is shown to be first-order accurate and singular perturbation parameter uniform convergence. Finally, numerical results are presented, which are in agreement with theoretical results.