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.

Dynamic visual features for audio-visual speaker verification

The cascading appearance-based (CAB) feature extraction technique has established itself as the state-of-the-art in extracting dynamic visual speech features for speech recognition. In this paper, we will focus on investigating the effectiveness of this technique for the related speaker verification application. By investigating the speaker verification ability of each stage of the cascade we will demonstrate that the same steps taken to reduce static speaker and environmental information for the visual speech recognition application also provide similar improvements for visual speaker recognition. A further study is conducted comparing synchronous HMM (SHMM) based fusion of CAB visual features and traditional perceptual linear predictive (PLP) acoustic features to show that higher complexity inherit in the SHMM approach does not appear to provide any improvement in the final audio-visual speaker verification system over simpler utterance level score fusion.

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.

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 Verification: Application task of EVALITA 2009. This competitive submission consisted of a score-level fusion of three component systems; a joint-factor analysis GMM system and two SVM systems using GLDS and GMM supervector kernels. Development evaluation and post-submission results are presented in this study, demonstrating the effectiveness of this fused system approach. This study highlights the challenges associated with system calibration from limited development data and that mismatch between training and testing conditions continues to be a major source of error in speaker verification technology.

Soundness of workflow nets with reset arcs

Petri nets are often used to model and analyze workflows. Many workflow languages have been mapped onto Petri nets in order to provide formal semantics or to verify correctness properties. Typically, the so-called Workflow nets are used to model and analyze workflows and variants of the classical soundness property are used as a correctness notion. Since many workflow languages have cancelation features, a mapping to workflow nets is not always possible. Therefore, it is interesting to consider workflow nets with reset arcs. Unfortunately, soundness is undecidable for workflow nets with reset arcs. In this paper, we provide a proof and insights into the theoretical limits of workflow verification.

Vector quantization based Gaussian modeling for speaker verification

Gaussian mixture models (GMMs) have become an established means of modeling feature distributions in speaker recognition systems. It is useful for experimentation and practical implementation purposes to develop and test these models in an efficient manner particularly when computational resources are limited. A method of combining vector quantization (VQ) with single multi-dimensional Gaussians is proposed to rapidly generate a robust model approximation to the Gaussian mixture model. A fast method of testing these systems is also proposed and implemented. Results on the NIST 1996 Speaker Recognition Database suggest comparable and in some cases an improved verification performance to the traditional GMM based analysis scheme. In addition, previous research for the task of speaker identification indicated a similar system perfomance between the VQ Gaussian based technique and GMMs

Robust speaker verification via fusion of speech and lip modalities

This paper investigates the use of lip information, in conjunction with speech information, for robust speaker verification in the presence of background noise. It has been previously shown in our own work, and in the work of others, that features extracted from a speaker's moving lips hold speaker dependencies which are complementary with speech features. We demonstrate that the fusion of lip and speech information allows for a highly robust speaker verification system which outperforms the performance of either sub-system. We present a new technique for determining the weighting to be applied to each modality so as to optimize the performance of the fused system. Given a correct weighting, lip information is shown to be highly effective for reducing the false acceptance and false rejection error rates in the presence of background noise

The use of speech and lip modalities for robust speaker verification under adverse conditions

Investigates the use of lip information, in conjunction with speech information, for robust speaker verification in the presence of background noise. We have previously shown (Int. Conf. on Acoustics, Speech and Signal Proc., vol. 6, pp. 3693-3696, May 1998) that features extracted from a speaker's moving lips hold speaker dependencies which are complementary with speech features. We demonstrate that the fusion of lip and speech information allows for a highly robust speaker verification system which outperforms either subsystem individually. We present a new technique for determining the weighting to be applied to each modality so as to optimize the performance of the fused system. Given a correct weighting, lip information is shown to be highly effective for reducing the false acceptance and false rejection error rates in the presence of background noise

QoS-aware web service composition using genetic algorithms

Web service technology is increasingly being used to build various e-Applications, in domains such as e-Business and e-Science. Characteristic benefits of web service technology are its inter-operability, decoupling and just-in-time integration. Using web service technology, an e-Application can be implemented by web service composition — by composing existing individual web services in accordance with the business process of the application. This means the application is provided to customers in the form of a value-added composite web service. An important and challenging issue of web service composition, is how to meet Quality-of-Service (QoS) requirements. This includes customer focused elements such as response time, price, throughput and reliability as well as how to best provide QoS results for the composites. This in turn best fulfils customers’ expectations and achieves their satisfaction. Fulfilling these QoS requirements or addressing the QoS-aware web service composition problem is the focus of this project. From a computational point of view, QoS-aware web service composition can be transformed into diverse optimisation problems. These problems are characterised as complex, large-scale, highly constrained and multi-objective problems. We therefore use genetic algorithms (GAs) to address QoS-based service composition problems. More precisely, this study addresses three important subproblems of QoS-aware web service composition; QoS-based web service selection for a composite web service accommodating constraints on inter-service dependence and conflict, QoS-based resource allocation and scheduling for multiple composite services on hybrid clouds, and performance-driven composite service partitioning for decentralised execution. Based on operations research theory, we model the three problems as a constrained optimisation problem, a resource allocation and scheduling problem, and a graph partitioning problem, respectively. Then, we present novel GAs to address these problems. We also conduct experiments to evaluate the performance of the new GAs. Finally, verification experiments are performed to show the correctness of the GAs. The major outcomes from the first problem are three novel GAs: a penaltybased GA, a min-conflict hill-climbing repairing GA, and a hybrid GA. These GAs adopt different constraint handling strategies to handle constraints on interservice dependence and conflict. This is an important factor that has been largely ignored by existing algorithms that might lead to the generation of infeasible composite services. Experimental results demonstrate the effectiveness of our GAs for handling the QoS-based web service selection problem with constraints on inter-service dependence and conflict, as well as their better scalability than the existing integer programming-based method for large scale web service selection problems. The major outcomes from the second problem has resulted in two GAs; a random-key GA and a cooperative coevolutionary GA (CCGA). Experiments demonstrate the good scalability of the two algorithms. In particular, the CCGA scales well as the number of composite services involved in a problem increases, while no other algorithms demonstrate this ability. The findings from the third problem result in a novel GA for composite service partitioning for decentralised execution. Compared with existing heuristic algorithms, the new GA is more suitable for a large-scale composite web service program partitioning problems. In addition, the GA outperforms existing heuristic algorithms, generating a better deployment topology for a composite web service for decentralised execution. These effective and scalable GAs can be integrated into QoS-based management tools to facilitate the delivery of feasible, reliable and high quality composite web services.

The development of Monte Carlo techniques for the verification of radiotherapy treatments

Radiotherapy is a cancer treatment modality in which a dose of ionising radiation is delivered to a tumour. The accurate calculation of the dose to the patient is very important in the design of an effective therapeutic strategy. This study aimed to systematically examine the accuracy of the radiotherapy dose calculations performed by clinical treatment planning systems by comparison againstMonte Carlo simulations of the treatment delivery. A suite of software tools known as MCDTK (Monte Carlo DICOM ToolKit) was developed for this purpose, and is capable of: • Importing DICOM-format radiotherapy treatment plans and producing Monte Carlo simulation input files (allowing simple simulation of complex treatments), and calibrating the results; • Analysing the predicted doses of and deviations between the Monte Carlo simulation results and treatment planning system calculations in regions of interest (tumours and organs-at-risk) and generating dose-volume histograms, so that conformity with dose prescriptions can be evaluated. The code has been tested against various treatment planning systems, linear acceleratormodels and treatment complexities. Six clinical head and neck cancer treatments were simulated and the results analysed using this software. The deviations were greatest where the treatment volume encompassed tissues on both sides of an air cavity. This was likely due to the method the planning system used to model low density media.