Bayes factor based speaker clustering for speaker diarization

This paper proposes the use of the Bayes Factor to replace the Bayesian Information Criterion (BIC) as a criterion for speaker clustering within a speaker diarization system. The BIC is one of the most popular decision criteria used in speaker diarization systems today. However, it will be shown in this paper that the BIC is only an approximation to the Bayes factor of marginal likelihoods of the data given each hypothesis. This paper uses the Bayes factor directly as a decision criterion for speaker clustering, thus removing the error introduced by the BIC approximation. Results obtained on the 2002 Rich Transcription (RT-02) Evaluation dataset show an improved clustering performance, leading to a 14.7% relative improvement in the overall Diarization Error Rate (DER) compared to the baseline system.

Yield-factor volatility models

The term structure of interest rates is often summarized using a handful of yield factors that capture shifts in the shape of the yield curve. In this paper, we develop a comprehensive model for volatility dynamics in the level, slope, and curvature of the yield curve that simultaneously includes level and GARCH effects along with regime shifts. We show that the level of the short rate is useful in modeling the volatility of the three yield factors and that there are significant GARCH effects present even after including a level effect. Further, we find that allowing for regime shifts in the factor volatilities dramatically improves the model’s fit and strengthens the level effect. We also show that a regime-switching model with level and GARCH effects provides the best out-of-sample forecasting performance of yield volatility. We argue that the auxiliary models often used to estimate term structure models with simulation-based estimation techniques should be consistent with the main features of the yield curve that are identified by our model.

Investigation of highly regulated promoters for the production of recombinant proteins in plants

Plants have been identified as promising expression systems for the commercial production of recombinant proteins. Plant-based protein production or “biofarming” offers a number of advantages over traditional expression systems in terms of scale of production, the capacity for post-translation processing, providing a product free of contaminants and cost effectiveness. A number of pharmaceutically important and commercially valuable proteins, such as antibodies, biopharmaceuticals and industrial enzymes are currently being produced in plant expression systems. However, several challenges still remain to improve recombinant protein yield with no ill effect on the host plant. The ability for transgenic plants to produce foreign proteins at commercially viable levels can be directly related to the level and cell specificity of the selected promoter driving the transgene. The accumulation of recombinant proteins may be controlled by a tissue-specific, developmentally-regulated or chemically-inducible promoter such that expression of recombinant proteins can be spatially- or temporally- controlled. The strict control of gene expression is particularly useful for proteins that are considered toxic and whose expression is likely to have a detrimental effect on plant growth. To date, the most commonly used promoter in plant biotechnology is the cauliflower mosaic virus (CaMV) 35S promoter which is used to drive strong, constitutive transgene expression in most organs of transgenic plants. Of particular interest to researchers in the Centre for Tropical Crops and Biocommodities at QUT are tissue-specific promoters for the accumulation of foreign proteins in the roots, seeds and fruit of various plant species, including tobacco, banana and sugarcane. Therefore this Masters project aimed to isolate and characterise root- and seed-specific promoters for the control of genes encoding recombinant proteins in plant-based expression systems. Additionally, the effects of matching cognate terminators with their respective gene promoters were assessed. The Arabidopsis root promoters ARSK1 and EIR1 were selected from the literature based on their reported limited root expression profiles. Both promoters were analysed using the PlantCARE database to identify putative motifs or cis-acting elements that may be associated with this activity. A number of motifs were identified in the ARSK1 promoter region including, WUN (wound-inducible), MBS (MYB binding site), Skn-1, and a RY core element (seed-specific) and in the EIR1 promoter region including, Skn-1 (seed-specific), Box-W1 (fungal elicitor), Aux-RR core (auxin response) and ABRE (ABA response). However, no previously reported root-specific cis-acting elements were observed in either promoter region. To confirm root specificity, both promoters, and truncated versions, were fused to the GUS reporter gene and the expression cassette introduced into Arabidopsis via Agrobacterium-mediated transformation. Despite the reported tissue-specific nature of these promoters, both upstream regulatory regions directed constitutive GUS expression in all transgenic plants. Further, similar levels of GUS expression from the ARSK1 promoter were directed by the control CaMV 35S promoter. The truncated version of the EIR1 promoter (1.2 Kb) showed some differences in the level of GUS expression compared to the 2.2 Kb promoter. Therefore, this suggests an enhancer element is contained in the 2.2 Kb upstream region that increases transgene expression. The Arabidopsis seed-specific genes ATS1 and ATS3 were selected from the literature based on their seed-specific expression profiles and gene expression confirmed in this study as seed-specific by RT-PCR analysis. The selected promoter regions were analysed using the PlantCARE database in order to identify any putative cis elements. The seed-specific motifs GCN4 and Skn-1 were identified in both promoter regions that are associated with elevated expression levels in the endosperm. Additionaly, the seed-specific RY element and the ABRE were located in the ATS1 promoter. Both promoters were fused to the GUS reporter gene and used to transform Arabidopsis plants. GUS expression from the putative promoters was consitutive in all transgenic Arabidopsis tissue tested. Importantly, the positive control FAE1 seed-specific promoter also directed constitutive GUS expression throughout transgenic Arabidopsis plants. The constitutive nature seen in all of the promoters used in this study was not anticipated. While variations in promoter activity can be caused by a number of influencing factors, the variation in promoter activity observed here would imply a major contributing factor common to all plant expression cassettes tested. All promoter constructs generated in this study were based on the binary vector pCAMBIA2300. This vector contains the plant selection gene (NPTII) under the transcriptional control of the duplicated CaMV 35S promoter. This CaMV 35S promoter contains two enhancer domains that confer strong, constitutive expression of the selection gene and is located immediately upstream of the promoter-GUS fusion. During the course of this project, Yoo et al. (2005) reported that transgene expression is significantly affected when the expression cassette is located on the same T-DNA as the 35S enhancer. It was concluded, the trans-acting effects of the enhancer activate and control transgene expression causing irregular expression patterns. This phenomenon seems the most plausible reason for the constitutive expression profiles observed with the root- and seed-specific promoters assessed in this study. The expression from some promoters can be influenced by their cognate terminator sequences. Therefore, the Arabidopsis ARSK1, EIR1, ATS1 and ATS3 terminator sequences were isolated and incorporated into expression cassettes containing the GUS reporter gene under the control of their cognate promoters. Again, unrestricted GUS activity was displayed throughout transgenic plants transformed with these reporter gene fusions. As previously discussed constitutive GUS expression was most likely due to the trans-acting effect of the upstream CaMV 35S promoter in the selection cassette located on the same T-DNA. The results obtained in this study make it impossible to assess the influence matching terminators with their cognate promoters have on transgene expression profiles. The obvious future direction of research continuing from this study would be to transform pBIN-based promoter-GUS fusions (ie. constructs containing no CaMV 35S promoter driving the plant selection gene) into Arabidopsis in order to determine the true tissue specificity of these promoters and evaluate the effects of their cognate 3’ terminator sequences. Further, promoter truncations based around the cis-elements identified here may assist in determining whether these motifs are in fact involved in the overall activity of the promoter.

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.

Brain-derived neurotrophic factor (BDNF) gene : no major impact on antidepressant treatment response

The brain-derived neurotrophic factor (BDNF) has been suggested to play a pivotal role in the aetiology of affective disorders. In order to further clarify the impact of BDNF gene variation on major depression as well as antidepressant treatment response, association of three BDNF polymorphisms [rs7103411, Val66Met (rs6265) and rs7124442] with major depression and antidepressant treatment response was investigated in an overall sample of 268 German patients with major depression and 424 healthy controls. False discovery rate (FDR) was applied to control for multiple testing. Additionally, ten markers in BDNF were tested for association with citalopram outcome in the STAR*D sample. While BDNF was not associated with major depression as a categorical diagnosis, the BDNF rs7124442 TT genotype was significantly related to worse treatment outcome over 6 wk in major depression (p=0.01) particularly in anxious depression (p=0.003) in the German sample. However, BDNF rs7103411 and rs6265 similarly predicted worse treatment response over 6 wk in clinical subtypes of depression such as melancholic depression only (rs7103411: TT

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.

Autocrine fibroblast growth factor 2 increases the multipotentiality of human adipose-derived mesenchymal stem cells

Multipotent mesenchymal stem cells (MSCs), first identified in the bone marrow, have subsequently been found in many other tissues, including fat, cartilage, muscle, and bone. Adipose tissue has been identified as an alternative to bone marrow as a source for the isolation of MSCs, as it is neither limited in volume nor as invasive in the harvesting. This study compares the multipotentiality of bone marrow-derived mesenchymal stem cells (BMSCs) with that of adipose-derived mesenchymal stem cells (AMSCs) from 12 age- and sex-matched donors. Phenotypically, the cells are very similar, with only three surface markers, CD106, CD146, and HLA-ABC, differentially expressed in the BMSCs. Although colony-forming units-fibroblastic numbers in BMSCs were higher than in AMSCs, the expression of multiple stem cell-related genes, like that of fibroblast growth factor 2 (FGF2), the Wnt pathway effectors FRAT1 and frizzled 1, and other self-renewal markers, was greater in AMSCs. Furthermore, AMSCs displayed enhanced osteogenic and adipogenic potential, whereas BMSCs formed chondrocytes more readily than AMSCs. However, by removing the effects of proliferation from the experiment, AMSCs no longer out-performed BMSCs in their ability to undergo osteogenic and adipogenic differentiation. Inhibition of the FGF2/fibroblast growth factor receptor 1 signaling pathway demonstrated that FGF2 is required for the proliferation of both AMSCs and BMSCs, yet blocking FGF2 signaling had no direct effect on osteogenic differentiation. Disclosure of potential conflicts of interest is found at the end of this article.

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.

A new future for growth factor complexes as wound therapies?

Background: Topical administration of growth factors (GFs) has displayed some potential in wound healing, but variable efficacy, high doses and costs have hampered their implementation. Moreover, this approach ignores the fact that wound repair is driven by interactions between multiple GFs and extracellular matrix (ECM) proteins. The Problem: Deep dermal partial thickness burn (DDPTB) injuries are the most common burn presentation to pediatric hospitals and also represent the most difficult burn injury to manage clinically. DDPTB often repair with a hypertrophic scar. Wounds that close rapidly exhibit reduced scarring. Thus treatments that shorten the time taken to close DDTPB’s may coincidently reduce scarring. Basic/Clinical Science Advances: We have observed that multi-protein complexes comprised of IGF and IGF-binding proteins bound to the ECM protein vitronectin (VN) significantly enhance cellular functions relevant to wound repair in human skin keratinocytes. These responses require activation of both the IGF-1R and the VN-binding αv integrins. We have recently evaluated the wound healing potential of these GF:VN complexes in a porcine model of DDTPB injury. Clinical Care Relevance: This pilot study demonstrates that GF:VN complexes hold promise as a wound healing therapy. Enhanced healing responses were observed after treatment with nanogram doses of the GF:VN complexes in vitro and in vivo. Critically healing was achieved using substantially less GF than studies in which GFs alone have been used. Conclusion: These data suggest that coupling GFs to ECM proteins, such as VN, may ultimately prove to be an improved technique for the delivery of novel GF-based wound therapies.

A confirmatory factor analysis of the Toronto Alexithymia Scale (TAS-20) in an alcohol-dependent sample

Confirmatory factor analyses were conducted to evaluate the factorial validity of the Toronto Alexithymia Scale in an alcohol-dependent sample. Several factor models were examined, but all models were rejected given their poor fit. A revision of the TAS-20 in alcohol-dependent populations may be needed.

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.

Fingerprint analysis using wavelet transform with application to compression and feature extraction

The main goal of this research is to design an efficient compression al~ gorithm for fingerprint images. The wavelet transform technique is the principal tool used to reduce interpixel redundancies and to obtain a parsimonious representation for these images. A specific fixed decomposition structure is designed to be used by the wavelet packet in order to save on the computation, transmission, and storage costs. This decomposition structure is based on analysis of information packing performance of several decompositions, two-dimensional power spectral density, effect of each frequency band on the reconstructed image, and the human visual sensitivities. This fixed structure is found to provide the "most" suitable representation for fingerprints, according to the chosen criteria. Different compression techniques are used for different subbands, based on their observed statistics. The decision is based on the effect of each subband on the reconstructed image according to the mean square criteria as well as the sensitivities in human vision. To design an efficient quantization algorithm, a precise model for distribution of the wavelet coefficients is developed. The model is based on the generalized Gaussian distribution. A least squares algorithm on a nonlinear function of the distribution model shape parameter is formulated to estimate the model parameters. A noise shaping bit allocation procedure is then used to assign the bit rate among subbands. To obtain high compression ratios, vector quantization is used. In this work, the lattice vector quantization (LVQ) is chosen because of its superior performance over other types of vector quantizers. The structure of a lattice quantizer is determined by its parameters known as truncation level and scaling factor. In lattice-based compression algorithms reported in the literature the lattice structure is commonly predetermined leading to a nonoptimized quantization approach. In this research, a new technique for determining the lattice parameters is proposed. In the lattice structure design, no assumption about the lattice parameters is made and no training and multi-quantizing is required. The design is based on minimizing the quantization distortion by adapting to the statistical characteristics of the source in each subimage. 11 Abstract Abstract Since LVQ is a multidimensional generalization of uniform quantizers, it produces minimum distortion for inputs with uniform distributions. In order to take advantage of the properties of LVQ and its fast implementation, while considering the i.i.d. nonuniform distribution of wavelet coefficients, the piecewise-uniform pyramid LVQ algorithm is proposed. The proposed algorithm quantizes almost all of source vectors without the need to project these on the lattice outermost shell, while it properly maintains a small codebook size. It also resolves the wedge region problem commonly encountered with sharply distributed random sources. These represent some of the drawbacks of the algorithm proposed by Barlaud [26). The proposed algorithm handles all types of lattices, not only the cubic lattices, as opposed to the algorithms developed by Fischer [29) and Jeong [42). Furthermore, no training and multiquantizing (to determine lattice parameters) is required, as opposed to Powell's algorithm [78). For coefficients with high-frequency content, the positive-negative mean algorithm is proposed to improve the resolution of reconstructed images. For coefficients with low-frequency content, a lossless predictive compression scheme is used to preserve the quality of reconstructed images. A method to reduce bit requirements of necessary side information is also introduced. Lossless entropy coding techniques are subsequently used to remove coding redundancy. The algorithms result in high quality reconstructed images with better compression ratios than other available algorithms. To evaluate the proposed algorithms their objective and subjective performance comparisons with other available techniques are presented. The quality of the reconstructed images is important for a reliable identification. Enhancement and feature extraction on the reconstructed images are also investigated in this research. A structural-based feature extraction algorithm is proposed in which the unique properties of fingerprint textures are used to enhance the images and improve the fidelity of their characteristic features. The ridges are extracted from enhanced grey-level foreground areas based on the local ridge dominant directions. The proposed ridge extraction algorithm, properly preserves the natural shape of grey-level ridges as well as precise locations of the features, as opposed to the ridge extraction algorithm in [81). Furthermore, it is fast and operates only on foreground regions, as opposed to the adaptive floating average thresholding process in [68). Spurious features are subsequently eliminated using the proposed post-processing scheme.

Spectral coding methods for speech compression and speaker identification

This thesis investigates aspects of encoding the speech spectrum at low bit rates, with extensions to the effect of such coding on automatic speaker identification. Vector quantization (VQ) is a technique for jointly quantizing a block of samples at once, in order to reduce the bit rate of a coding system. The major drawback in using VQ is the complexity of the encoder. Recent research has indicated the potential applicability of the VQ method to speech when product code vector quantization (PCVQ) techniques are utilized. The focus of this research is the efficient representation, calculation and utilization of the speech model as stored in the PCVQ codebook. In this thesis, several VQ approaches are evaluated, and the efficacy of two training algorithms is compared experimentally. It is then shown that these productcode vector quantization algorithms may be augmented with lossless compression algorithms, thus yielding an improved overall compression rate. An approach using a statistical model for the vector codebook indices for subsequent lossless compression is introduced. This coupling of lossy compression and lossless compression enables further compression gain. It is demonstrated that this approach is able to reduce the bit rate requirement from the current 24 bits per 20 millisecond frame to below 20, using a standard spectral distortion metric for comparison. Several fast-search VQ methods for use in speech spectrum coding have been evaluated. The usefulness of fast-search algorithms is highly dependent upon the source characteristics and, although previous research has been undertaken for coding of images using VQ codebooks trained with the source samples directly, the product-code structured codebooks for speech spectrum quantization place new constraints on the search methodology. The second major focus of the research is an investigation of the effect of lowrate spectral compression methods on the task of automatic speaker identification. The motivation for this aspect of the research arose from a need to simultaneously preserve the speech quality and intelligibility and to provide for machine-based automatic speaker recognition using the compressed speech. This is important because there are several emerging applications of speaker identification where compressed speech is involved. Examples include mobile communications where the speech has been highly compressed, or where a database of speech material has been assembled and stored in compressed form. Although these two application areas have the same objective - that of maximizing the identification rate - the starting points are quite different. On the one hand, the speech material used for training the identification algorithm may or may not be available in compressed form. On the other hand, the new test material on which identification is to be based may only be available in compressed form. Using the spectral parameters which have been stored in compressed form, two main classes of speaker identification algorithm are examined. Some studies have been conducted in the past on bandwidth-limited speaker identification, but the use of short-term spectral compression deserves separate investigation. Combining the major aspects of the research, some important design guidelines for the construction of an identification model when based on the use of compressed speech are put forward.