Acoustic keyword spotting in speech with applications to data mining

Keyword Spotting is the task of detecting keywords of interest within continu- ous speech. The applications of this technology range from call centre dialogue systems to covert speech surveillance devices. Keyword spotting is particularly well suited to data mining tasks such as real-time keyword monitoring and unre- stricted vocabulary audio document indexing. However, to date, many keyword spotting approaches have su®ered from poor detection rates, high false alarm rates, or slow execution times, thus reducing their commercial viability. This work investigates the application of keyword spotting to data mining tasks. The thesis makes a number of major contributions to the ¯eld of keyword spotting. The ¯rst major contribution is the development of a novel keyword veri¯cation method named Cohort Word Veri¯cation. This method combines high level lin- guistic information with cohort-based veri¯cation techniques to obtain dramatic improvements in veri¯cation performance, in particular for the problematic short duration target word class. The second major contribution is the development of a novel audio document indexing technique named Dynamic Match Lattice Spotting. This technique aug- ments lattice-based audio indexing principles with dynamic sequence matching techniques to provide robustness to erroneous lattice realisations. The resulting algorithm obtains signi¯cant improvement in detection rate over lattice-based audio document indexing while still maintaining extremely fast search speeds. The third major contribution is the study of multiple veri¯er fusion for the task of keyword veri¯cation. The reported experiments demonstrate that substantial improvements in veri¯cation performance can be obtained through the fusion of multiple keyword veri¯ers. The research focuses on combinations of speech background model based veri¯ers and cohort word veri¯ers. The ¯nal major contribution is a comprehensive study of the e®ects of limited training data for keyword spotting. This study is performed with consideration as to how these e®ects impact the immediate development and deployment of speech technologies for non-English languages.

Semantic search and inferencing in health informatics

Consider a person searching electronic health records, a search for the term ‘cracked skull’ should return documents that contain the term ‘cranium fracture’. A information retrieval systems is required that matches concepts, not just keywords. Further more, determining relevance of a query to a document requires inference – its not simply matching concepts. For example a document containing ‘dialysis machine’ should align with a query for ‘kidney disease’. Collectively we describe this problem as the ‘semantic gap’ – the difference between the raw medical data and the way a human interprets it. This paper presents an approach to semantic search of health records by combining two previous approaches: an ontological approach using the SNOMED CT medical ontology; and a distributional approach using semantic space vector space models. Our approach will be applied to a specific problem in health informatics: the matching of electronic patient records to clinical trials.

Designing participation : Reconsidering agile ridesharing with evolutionary distributed design thinking

This paper explores design thinking from the perspective of designing new forms of interaction to engage people in community change initiatives. A case study of an agile ridesharing system is presented. We describe the fundamental premise of the design approach taken—deploying simple interactive prototypes for use by communities in order to test the design hypothesis, evolve the design in use and grow the community of participants. Real-time use data and feedback from participants influences our understanding of the design approach and feeds into the gradual evolution of the prototype while it continues to be used. We then reflect upon this form of evolutionary distributed design thinking. In contrast to the conventional IT wisdom of building systems to automate ride matching and fare calculation using structured forms, our initial phase of design revealed a preference for informal messaging, negotiation and caution in the sharing of specific location information.

Cystine-mediated oxidative defence in Lactobacillus reuteri BR11

Lactobacillus reuteri BR11 possesses an abundant cystine uptake (Cyu) ABC-transporter that was previously found to be involved in a novel mechanism of oxidative defence mediated by cystine. The current study aimed to elucidate this mechanism with a focus on the role of the co-transcribed cystathionine ã-lyase (Cgl). Growth studies of wild-type L. reuteri BR11 and mutants inactivated in cgl and the cystine-binding protein encoding gene cyuC showed that in contrast to the Cyu transporter, whose inactivation led to growth arrest in aerated cultures, Cgl is not crucial for oxidative defence. However, the role of Cgl in oxidative defence became apparent in the presence of severe oxidative damage and cysteine deprivation. Cysteine was found to be protective against oxidative stress, and the action of Cgl in both cysteine biosynthesis and degradation poses a seemingly futile pathway that deprives the intracellular cysteine pool. To further characterise the relationship between Cgl activity and cysteine and their roles in oxidative defence, enzymatic assays were performed on purified Cgl, and intracellular concentrations of cysteine, cystathionine and methionine were determined. Cgl was highly active towards cystine and cystathionine and less active towards cysteine in vitro, suggesting the main function of Cgl to be cysteine biosynthesis. Cysteine was found at high concentrations in the cell, but the levels were not significantly affected by inactivation of cgl or growth under aerobic conditions. It was concluded that both anabolic and catabolic activities of Cgl towards cysteine contribute to oxidative defence, the former by maintaining an intracellular reservoir of thiol analogous to glutathione, and the latter by producing H2S which is readily secreted, thus creating a reducing extracellular environment. The significance of the Cyu transporter to the physiology of L. reuteri BR11 prompted a phylogenetic study to determine its presence in bacteria. Orthologs of the Cyu transporter that are closest matches to the Cyu transporter are only limited to several species of Lactobacillus and Leuconostoc. Outside the Lactobacillales order, the closest matching orthologs belong to Proteobacteria, and there are more orthologs in Proteobacteria than non-Lactobacillales Firmicutes, suggesting that the Cyu transporter locus was present in the ancestor of the Proteobacteria and Firmicutes, and over evolutionary time has been lost or diverged in many Firmicutes. The clustering of the Cyu transporter locus with a gene encoding a Cgl family protein is even rarer. It was only found in L. reuteri, Lactobacillus vaginalis, Weissella paramesenteroides, the Lactobacillus casei group, and several Campylobacter sp. An accompanying phylogenetic study of L. reuteri BR11 using multi-locus sequence analysis showed that L. reuteri BR11 had diverged from more than 100 strains of L. reuteri isolated from various hosts and geographical locations. However, comparison with other Lactobacillus species supported the current classification of BR11 as L. reuteri. The most closely related species to L. reuteri is L. vaginalis or Lactobacillus antri, depending on the housekeeping gene used for analysis. The close evolutionary relationship of L. vaginalis to L. reuteri and the high degree of sequence identity between the cgl-cyuABC loci in both species suggest that the Cyu system is highly likely to perform similar functions in L. vaginalis. In search of other genes that function in oxidative defence, a number of mutants which were inactivated in genes that confer increased resistance to oxidative stress in other bacteria were constructed. The genes targeted were ahpC (peroxidase component of the alkyl hydroperoxide reductase system), tpx (thiol peroxidase), osmC (osmotically induced protein C), mntH (Mn2+/Fe2+ transporter), gshA (ã-glutamylcysteine synthetase) and msrA (methionine sulfoxide reductase). The ahpC and mntH mutants had slightly lower minimum inhibitory concentrations of organic peroxides, suggesting these genes might be involved in resistance to organic peroxides in L. reuteri. However, none of the mutants exhibited growth defects in aerated cultures, in stark contrast to the cyuC mutant. This may be due to compensatory functions of other genes, a hypothesis which cannot be tested until a robust protocol for constructing markerless multiple gene deletion mutants in L. reuteri is developed. These results highlight the importance of the Cyu transporter in oxidative defence and provide a foundation for extending the research of this system in other bacteria.

Fast and accurate phonetic spoken term detection

For the first time in human history, large volumes of spoken audio are being broadcast, made available on the internet, archived, and monitored for surveillance every day. New technologies are urgently required to unlock these vast and powerful stores of information. Spoken Term Detection (STD) systems provide access to speech collections by detecting individual occurrences of specified search terms. The aim of this work is to develop improved STD solutions based on phonetic indexing. In particular, this work aims to develop phonetic STD systems for applications that require open-vocabulary search, fast indexing and search speeds, and accurate term detection. Within this scope, novel contributions are made within two research themes, that is, accommodating phone recognition errors and, secondly, modelling uncertainty with probabilistic scores. A state-of-the-art Dynamic Match Lattice Spotting (DMLS) system is used to address the problem of accommodating phone recognition errors with approximate phone sequence matching. Extensive experimentation on the use of DMLS is carried out and a number of novel enhancements are developed that provide for faster indexing, faster search, and improved accuracy. Firstly, a novel comparison of methods for deriving a phone error cost model is presented to improve STD accuracy, resulting in up to a 33% improvement in the Figure of Merit. A method is also presented for drastically increasing the speed of DMLS search by at least an order of magnitude with no loss in search accuracy. An investigation is then presented of the effects of increasing indexing speed for DMLS, by using simpler modelling during phone decoding, with results highlighting the trade-off between indexing speed, search speed and search accuracy. The Figure of Merit is further improved by up to 25% using a novel proposal to utilise word-level language modelling during DMLS indexing. Analysis shows that this use of language modelling can, however, be unhelpful or even disadvantageous for terms with a very low language model probability. The DMLS approach to STD involves generating an index of phone sequences using phone recognition. An alternative approach to phonetic STD is also investigated that instead indexes probabilistic acoustic scores in the form of a posterior-feature matrix. A state-of-the-art system is described and its use for STD is explored through several experiments on spontaneous conversational telephone speech. A novel technique and framework is proposed for discriminatively training such a system to directly maximise the Figure of Merit. This results in a 13% improvement in the Figure of Merit on held-out data. The framework is also found to be particularly useful for index compression in conjunction with the proposed optimisation technique, providing for a substantial index compression factor in addition to an overall gain in the Figure of Merit. These contributions significantly advance the state-of-the-art in phonetic STD, by improving the utility of such systems in a wide range of applications.

Simulated cataract and low contrast stimuli impair cognitive performance in older adults : implications for neuropsychological assessment and everyday function

Objective: To investigate how age-related declines in vision (particularly contrast sensitivity), simulated using cataract-goggles and low-contrast stimuli, influence the accuracy and speed of cognitive test performance in older adults. An additional aim was to investigate whether declines in vision differentially affect secondary more than primary memory. Method: Using a fully within-subjects design, 50 older drivers aged 66-87 years completed two tests of cognitive performance - letter matching (perceptual speed) and symbol recall (short-term memory) - under different viewing conditions that degraded visual input (low-contrast stimuli, cataract-goggles, and low-contrast stimuli combined with cataract-goggles, compared with normal viewing). However, presentation time was also manipulated for letter matching. Visual function, as measured using standard charts, was taken into account in statistical analyses. Results: Accuracy and speed for cognitive tasks were significantly impaired when visual input was degraded. Furthermore, cognitive performance was positively associated with contrast sensitivity. Presentation time did not influence cognitive performance, and visual gradation did not differentially influence primary and secondary memory. Conclusion: Age-related declines in visual function can impact on the accuracy and speed of cognitive performance, and therefore the cognitive abilities of older adults may be underestimated in neuropsychological testing. It is thus critical that visual function be assessed prior to testing, and that stimuli be adapted to older adults' sensory capabilities (e.g., by maximising stimuli contrast).

The Culture of Older Women’s Drinking in Australia

The purpose of this investigation was to undertake pilot research to develop an understanding of the current culture of older Australian women’s (35-50 years) drinking behaviour from a uniquely female perspective. Methods Two separate focus group interviews were undertaken with women (N = 11) aged between 35 and 50 years living in South-East Queensland, Australia. Women were asked to openly discuss how and why they drink alcohol (ie., their regular drinking behaviour), how this has changed over time, and the attitudes and values that influence their behaviour. Results Participants reported that their consumption of alcohol was more regulated and controlled and although some women drank more frequently, the quantity consumed at each drinking occasion had decreased significantly. Occasional consumption of large amounts of alcohol tended to be the result of ‘incidental drinking’ as opposed to ‘determined drinking’. The reasons for alcohol consumption were found to be internal as well as social. Internal reasons included stress relief, increased relaxation and self reward. Further, alcohol was used as a social lubricant. This cohort also reported being influenced by the drinking patterns of their partners. Social group matching was however found to have a negative impact on alcohol consumption as social groups most commonly endorsed lesser levels of intoxication. Further, the women reported that they were of an age in which they felt excessive drinking to be ‘undignified’. Personal reasons such as vocational and family responsibilities further modified the levels of consumption for individual women. Finally, it was reported that perceived health risks that can result from excessive and/or repetitive drinking led to a decreased in consumption. Conclusion It is proposed that the findings of this investigation could be used to improve current knowledge regarding more mature women’s drinking culture, associated risks and risk prevention strategies.

A distributed protocol for object tracking in wireless multimedia sensor networks

We consider the problem of object tracking in a wireless multimedia sensor network (we mainly focus on the camera component in this work). The vast majority of current object tracking techniques, either centralised or distributed, assume unlimited energy, meaning these techniques don't translate well when applied within the constraints of low-power distributed systems. In this paper we develop and analyse a highly-scalable, distributed strategy to object tracking in wireless camera networks with limited resources. In the proposed system, cameras transmit descriptions of objects to a subset of neighbours, determined using a predictive forwarding strategy. The received descriptions are then matched at the next camera on the objects path using a probability maximisation process with locally generated descriptions. We show, via simulation, that our predictive forwarding and probabilistic matching strategy can significantly reduce the number of object-misses, ID-switches and ID-losses; it can also reduce the number of required transmissions over a simple broadcast scenario by up to 67%. We show that our system performs well under realistic assumptions about matching objects appearance using colour.

Personalizing the design of computer-based instruction to enhance learning

This paper reports two studies designed to investigate the effect on learning outcomes of matching individuals' preferred cognitive styles to computer-based instructional (CBI) material. Study 1 considered the styles individually as Verbalizer, Imager, Wholist and Analytic. Study 2 considered the bi-dimensional nature of cognitive styles in order to assess the full ramification of cognitive styles on learning: Analytic/Imager, Analytic/ Verbalizer, Wholist/Imager and the Wholist/Verbalizer. The mix of images and text, the nature of the text material, use of advance organizers and proximity of information to facilitate meaningful connections between various pieces of information were some of the considerations in the design of the CBI material. In a quasi-experimental format, students' cognitive styles were analysed by Cognitive Style Analysis (CSA) software. On the basis of the CSA result, the system defaulted students to either matched or mismatched CBI material by alternating between the two formats. The instructional material had a learning and a test phase. Learning outcome was tested on recall, labelling, explanation and problem-solving tasks. Comparison of the matched and mismatched instruction did not indicate significant difference between the groups, but the consistently better performance by the matched group suggests potential for further investigations where the limitations cited in this paper are eliminated. The result did indicate a significant difference between the four cognitive styles with the Wholist/Verbalizer group performing better then all other cognitive styles. Analysing the difference between cognitive styles on individual test tasks indicated significant difference on recall, labelling and explanation, suggesting that certain test tasks may suit certain cognitive styles.

Toward a more robust facial expression recognition in occluded images using randomly sampled Gabor based templates

Occlusion is a big challenge for facial expression recognition (FER) in real-world situations. Previous FER efforts to address occlusion suffer from loss of appearance features and are largely limited to a few occlusion types and single testing strategy. This paper presents a robust approach for FER in occluded images and addresses these issues. A set of Gabor based templates is extracted from images in the gallery using a Monte Carlo algorithm. These templates are converted into distance features using template matching. The resulting feature vectors are robust to occlusion. Occluded eyes and mouth regions and randomly places occlusion patches are used for testing. Two testing strategies analyze the effects of these occlusions on the overall recognition performance as well as each facial expression. Experimental results on the Cohn-Kanade database confirm the high robustness of our approach and provide useful insights about the effects of occlusion on FER. Performance is also compared with previous approaches.

Clone detection in repositories of business process models

As organizations reach to higher levels of business process management maturity, they often find themselves maintaining repositories of hundreds or even thousands of process models, representing valuable knowledge about their operations. Over time, process model repositories tend to accumulate duplicate fragments (also called clones) as new process models are created or extended by copying and merging fragments from other models. This calls for methods to detect clones in process models, so that these clones can be refactored as separate subprocesses in order to improve maintainability. This paper presents an indexing structure to support the fast detection of clones in large process model repositories. The proposed index is based on a novel combination of a method for process model decomposition (specifically the Refined Process Structure Tree), with established graph canonization and string matching techniques. Experiments show that the algorithm scales to repositories with hundreds of models. The experimental results also show that a significant number of non-trivial clones can be found in process model repositories taken from industrial practice.

Patient-specific computational biomechanics for simulating adolescent scoliosis surgery : predicted vs clinical correction for a preliminary series of six patients

Scoliosis is a spinal deformity that requires surgical correction in progressive cases. In order to optimize surgical outcomes, patient-specific finite element models are being developed by our group. In this paper, a single rod anterior correction procedure is simulated for a group of six scoliosis patients. For each patient, personalised model geometry was derived from low-dose CT scans, and clinically measured intra-operative corrective forces were applied. However, tissue material properties were not patient-specific, being derived from existing literature. Clinically, the patient group had a mean initial Cobb angle of 47.3 degrees, which was corrected to 17.5 degrees after surgery. The mean simulated post-operative Cobb angle for the group was 18.1 degrees. Although this represents good agreement between clinical and simulated corrections, the discrepancy between clinical and simulated Cobb angle for individual patients varied between -10.3 and +8.6 degrees, with only three of the six patients matching the clinical result to within accepted Cobb measurement error of +-5 degrees. The results of this study suggest that spinal tissue material properties play an important role in governing the correction obtained during surgery, and that patient-specific modelling approaches must address the question of how to prescribe patient-specific soft tissue properties for spine surgery simulation.

Early childhood teachers' mathematical content knowledge

The process of becoming numerate begins in the early years. According to Vygotskian theory (1978), teachers are More Knowledgeable Others who provide and support learning experiences that influence children’s mathematical learning. This paper reports on research that investigates three early childhood teachers mathematics content knowledge. An exploratory, single case study utilised data collected from interviews, and email correspondence to investigate the teachers’ mathematics content knowledge. The data was reviewed according to three analytical strategies: content analysis, pattern matching, and comparative analysis. Findings indicated there was variation in teachers’ content knowledge across the five mathematical strands and that teachers might not demonstrate the depth of content knowledge that is expected of four year specially trained early years’ teachers. A significant factor that appeared to influence these teachers’ content knowledge was their teaching experience. Therefore, an avenue for future research is the investigation of factors that influence teachers’ content numeracy knowledge.

Panel studies of new venture creation: a methods-focused review and suggestions for future research

Longitudinal panel studies of large, random samples of business start-ups captured at the pre-operational stage allow researchers to address core issues for entrepreneurship research, namely, the processes of creation of new business ventures as well as their antecedents and outcomes. Here, we perform a methods-orientated review of all 83 journal articles that have used this type of data set, our purpose being to assist users of current data sets as well as designers of new projects in making the best use of this innovative research approach. Our review reveals a number of methods issues that are largely particular to this type of research. We conclude that amidst exemplary contributions, much of the reviewed research has not adequately managed these methods challenges, nor has it made use of the full potential of this new research approach. Specifically, we identify and suggest remedies for context-specific and interrelated methods challenges relating to sample definition, choice of level of analysis, operationalization and conceptualization, use of longitudinal data and dealing with various types of problematic heterogeneity. In addition, we note that future research can make further strides towards full utilization of the advantages of the research approach through better matching (from either direction) between theories and the phenomena captured in the data, and by addressing some under-explored research questions for which the approach may be particularly fruitful.

Object identification based on 3D spatial models of construction sites

On obstacle-cluttered construction sites, understanding the motion characteristics of objects is important for anticipating collisions and preventing accidents. This study investigates algorithms for object identification applications that can be used by heavy equipment operators to effectively monitor congested local environment. The proposed framework contains algorithms for three-dimensional spatial modeling and image matching that are based on 3D images scanned by a high-frame rate range sensor. The preliminary results show that an occupancy grid spatial modeling algorithm can successfully build the most pertinent spatial information, and that an image matching algorithm is best able to identify which objects are in the scanned scene.