Improved facial expression recognition via uni-hyperplane classification

Large margin learning approaches, such as support vector machines (SVM), have been successfully applied to numerous classification tasks, especially for automatic facial expression recognition. The risk of such approaches however, is their sensitivity to large margin losses due to the influence from noisy training examples and outliers which is a common problem in the area of affective computing (i.e., manual coding at the frame level is tedious so coarse labels are normally assigned). In this paper, we leverage the relaxation of the parallel-hyperplanes constraint and propose the use of modified correlation filters (MCF). The MCF is similar in spirit to SVMs and correlation filters, but with the key difference of optimizing only a single hyperplane. We demonstrate the superiority of MCF over current techniques on a battery of experiments.

Hospital outpatient responses to potential driving impairment due to prescribed medications

Driving and using prescription medicines that have the potential to impair driving is an emerging research area. To date it is characterised by a limited (although growing) number of studies and methodological complexities that make generalisations about impairment due to medications difficult. Consistent evidence has been found for the impairing effects of hypnotics, sedative antidepressants and antihistamines, and narcotic analgesics, although it has been estimated that as many as nine medication classes have the potential to impair driving (Alvarez & del Rio, 2000; Walsh, de Gier, Christopherson, & Verstraete, 2004). There is also evidence for increased negative effects related to concomitant use of other medications and alcohol (Movig et al., 2004; Pringle, Ahern, Heller, Gold, & Brown, 2005). Statistics on the high levels of Australian prescription medication use suggest that consumer awareness of driving impairment due to medicines should be examined. One web-based study has found a low level of awareness, knowledge and risk perceptions among Australian drivers about the impairing effects of various medications on driving (Mallick, Johnston, Goren, & Kennedy, 2007). The lack of awareness and knowledge brings into question the effectiveness of the existing countermeasures. In Australia these consist of the use of ancillary warning labels administered under mandatory regulation and professional guidelines, advice to patients, and the use of Consumer Medicines Information (CMI) with medications that are known to cause impairment. The responsibility for the use of the warnings and related counsel to patients primarily lies with the pharmacist when dispensing relevant medication. A review by the Therapeutic Goods Administration (TGA) noted that in practice, advice to patients may not occur and that CMI is not always available (TGA, 2002). Researchers have also found that patients' recall of verbal counsel is very low (Houts, Bachrach, Witmer, Tringali, Bucher, & Localio, 1998). With healthcare observed as increasingly being provided in outpatient conditions (Davis et al., 2006; Vingilis & MacDonald, 2000), establishing the effectiveness of the warning labels as a countermeasure is especially important. There have been recent international developments in medication categorisation systems and associated medication warning labels. In 2005, France implemented a four-tier medication categorisation and warning system to improve patients' and health professionals' awareness and knowledge of related road safety issues (AFSSAPS, 2005). This warning system uses a pictogram and indicates the level of potential impairment in relation to driving performance through the use of colour and advice on the recommended behaviour to adopt towards driving. The comparable Australian system does not indicate the severity level of potential effects, and does not provide specific guidelines on the attitude or actions that the individual should adopt towards driving. It is reliant upon the patient to be vigilant in self-monitoring effects, to understand the potential ways in which they may be affected and how serious these effects may be, and to adopt the appropriate protective actions. This thesis investigates the responses of a sample of Australian hospital outpatients who receive appropriate labelling and counselling advice about potential driving impairment due to prescribed medicines. It aims to provide baseline data on the understanding and use of relevant medications by a Queensland public hospital outpatient sample recruited through the hospital pharmacy. It includes an exploration and comparison of the effect of the Australian and French medication warning systems on medication user knowledge, attitudes, beliefs and behaviour, and explores whether there are areas in which the Australian system may be improved by including any beneficial elements of the French system. A total of 358 outpatients were surveyed, and a follow-up telephone survey was conducted with a subgroup of consenting participants who were taking at least one medication that required an ancillary warning label about driving impairment. A complementary study of 75 French hospital outpatients was also conducted to further investigate the performance of the warnings. Not surprisingly, medication use among the Australian outpatient sample was high. The ancillary warning labels required to appear on medications that can impair driving were prevalent. A subgroup of participants was identified as being potentially at-risk of driving impaired, based on their reported recent use of medications requiring an ancillary warning label and level of driving activity. The sample reported previous behaviour and held future intentions that were consistent with warning label advice and health protective action. Participants did not express a particular need for being advised by a health professional regarding fitness to drive in relation to their medication. However, it was also apparent from the analysis that the participants would be significantly more likely to follow advice from a doctor than a pharmacist. High levels of knowledge in terms of general principles about effects of alcohol, illicit drugs and combinations of substances, and related health and crash risks were revealed. This may reflect a sample specific effect. Emphasis is placed in the professional guidelines for hospital pharmacists that make it essential that advisory labels are applied to medicines where applicable and that warning advice is given to all patients on medication which may affect driving (SHPA, 2006, p. 221). The research program applied selected theoretical constructs from Schwarzer's (1992) Health Action Process Approach, which has extended constructs from existing health theories such as the Theory of Planned Behavior (Ajzen, 1991) to better account for the intention-behaviour gap often observed when predicting behaviour. This was undertaken to explore the utility of the constructs in understanding and predicting compliance intentions and behaviour with the mandatory medication warning about driving impairment. This investigation revealed that the theoretical constructs related to intention and planning to avoid driving if an effect from the medication was noticed were useful. Not all the theoretical model constructs that had been demonstrated to be significant predictors in previous research on different health behaviours were significant in the present analyses. Positive outcome expectancies from avoiding driving were found to be important influences on forming the intention to avoid driving if an effect due to medication was noticed. In turn, intention was found to be a significant predictor of planning. Other selected theoretical constructs failed to predict compliance with the Australian warning label advice. It is possible that the limited predictive power of a number of constructs including risk perceptions is due to the small sample size obtained at follow up on which the evaluation is based. Alternately, it is possible that the theoretical constructs failed to sufficiently account for issues of particular relevance to the driving situation. The responses of the Australian hospital outpatient sample towards the Australian and French medication warning labels, which differed according to visual characteristics and warning message, were examined. In addition, a complementary study with a sample of French hospital outpatients was undertaken in order to allow general comparisons concerning the performance of the warnings. While a large amount of research exists concerning warning effectiveness, there is little research that has specifically investigated medication warnings relating to driving impairment. General established principles concerning factors that have been demonstrated to enhance warning noticeability and behavioural compliance have been extrapolated and investigated in the present study. The extent to which there is a need for education and improved health messages on this issue was a core issue of investigation in this thesis. Among the Australian sample, the size of the warning label and text, and red colour were the most visually important characteristics. The pictogram used in the French labels was also rated highly, and was salient for a large proportion of the sample. According to the study of French hospital outpatients, the pictogram was perceived to be the most important visual characteristic. Overall, the findings suggest that the Australian approach of using a combination of visual characteristics was important for the majority of the sample but that the use of a pictogram could enhance effects. A high rate of warning recall was found overall and a further important finding was that higher warning label recall was associated with increased number of medication classes taken. These results suggest that increased vigilance and care are associated with the number of medications taken and the associated repetition of the warning message. Significantly higher levels of risk perception were found for the French Level 3 (highest severity) label compared with the comparable mandatory Australian ancillary Label 1 warning. Participants' intentions related to the warning labels indicated that they would be more cautious while taking potentially impairing medication displaying the French Level 3 label compared with the Australian Label 1. These are potentially important findings for the Australian context regarding the current driving impairment warnings about displayed on medication. The findings raise other important implications for the Australian labelling context. An underlying factor may be the differences in the wording of the warning messages that appear on the Australian and French labels. The French label explicitly states "do not drive" while the Australian label states "if affected, do not drive", and the difference in responses may reflect that less severity is perceived where the situation involves the consumer's self-assessment of their impairment. The differences in the assignment of responsibility by the Australian (the consumer assesses and decides) and French (the doctor assesses and decides) approaches for the decision to drive while taking medication raises the core question of who is most able to assess driving impairment due to medication: the consumer, or the health professional? There are pros and cons related to knowledge, expertise and practicalities with either option. However, if the safety of the consumer is the primary aim, then the trend towards stronger risk perceptions and more consistent and cautious behavioural intentions in relation to the French label suggests that this approach may be more beneficial for consumer safety. The observations from the follow-up survey, although based on a small sample size and descriptive in nature, revealed that just over half of the sample recalled seeing a warning label about driving impairment on at least one of their medications. The majority of these respondents reported compliance with the warning advice. However, the results indicated variation in responses concerning alcohol intake and modifying the dose of medication or driving habits so that they could continue to drive, which suggests that the warning advice may not be having the desired impact. The findings of this research have implications for current countermeasures in this area. These have included enhancing the role that prescribing doctors have in providing warnings and advice to patients about the impact that their medication can have on driving, increasing consumer perceptions of the authority of pharmacists on this issue, and the reinforcement of the warning message. More broadly, it is suggested that there would be benefit in a wider dissemination of research-based information on increased crash risk and systematic monitoring and publicity about the representation of medications in crashes resulting in injuries and fatalities. Suggestions for future research concern the continued investigation of the effects of medications and interactions with existing medical conditions and other substances on driving skills, effects of variations in warning label design, individual behaviours and characteristics (particularly among those groups who are dependent upon prescription medication) and validation of consumer self-assessment of impairment.

Digital archives and cultural memory : discovering lost histories in digitised Australian children's literature 1851-1945

A recent Australian literature digitisation project uncovered some surprising discoveries in the children’s books that it digitised. The Children’s Literature Digital Resources (CLDR) Project digitised children’s books that were first published between 1851 to 1945 and made them available online through AustLit: The Australian Literature Resource. The digitisation process also preserved, within the pages of those books, a range of bookplates, book labels, inscriptions, and loose ephemera. This material allows us to trace the provenance of some of the digitised works, some of which came from the personal libraries of now-famous authors, and others from less celebrated sources. These extra-textual traces can contribute to cultural memory of the past by providing evidence of how books were collected and exchanged, and what kinds of books were presented as prizes in schools and Sunday schools. They also provide insight into Australian literary and artistic networks, particularly of the first few decades of the 20th century. This article describes the kinds of material uncovered in the digitisation process and suggests that the material provides insights into literary and cultural histories that might otherwise be forgotten. It also argues that the indexing of this material is vital if it is not to be lost to future researchers.

Naming or creating a problem?

In this chapter, we are going to consider how language and practice interact in the process of supporting the learning of students with diverse abilities. You will learn that it is necessary for teachers to understand that while labels carry an administrative function in schools, when used carelessly they operate to stigmatise and exclude those whom we are working to include. This chapter will introduce the concept of equity and explain how the dilemma of difference emerges when we try to determine who should receive support and how. The chapter will also explain how an appreciation of language can help to inform and transform our pedagogy. An example of inclusion in action is provided to illustrate how inclusive language in practice can promote deep cultural changes that benefit both students and teachers. The process of determining appropriate and effective education of students with additional support requirements is troubled by what some refer to as the ‘dilemma of difference’. This dilemma derives mainly from the nature of language and our need to use certain words, terms and categories in order to share common understandings. Without these, educators cannot hope to arrive on the same page, yet such words can take on a life of their own; influencing thoughts, perspectives and attitudes in ways that far outstrip original intentions. The drive for clarity, however, through definition and diagnostic classification can ultimately obscure because of the cultural meanings that become invested within these terms through their use over time and in different professional contexts. In effect, trying to define “difference” in order to provide the right support to particular students is a process that entrenches normative boundaries that in turn create, accentuate and stigmatise whatever we have decided constitutes difference. Language is thus a powerful and dangerous weapon but, like other weapons, language can both hurt and defend. Understanding the power of language enables educators to use it both wisely and safely to the maximum benefit of their students. This chapter will discuss how teachers can recognise and support their students in ways that avoid stigma and the closure of stereotyping.

Efficient querying of large process model repositories

Recent years have seen an increased uptake of business process management technology in industries. This has resulted in organizations trying to manage large collections of business process models. One of the challenges facing these organizations concerns the retrieval of models from large business process model repositories. For example, in some cases new process models may be derived from existing models, thus finding these models and adapting them may be more effective and less error-prone than developing them from scratch. Since process model repositories may be large, query evaluation may be time consuming. Hence, we investigate the use of indexes to speed up this evaluation process. To make our approach more applicable, we consider the semantic similarity between labels. Experiments are conducted to demonstrate that our approach is efficient.

Speaker diarization : "who spoke when"

Speaker diarization is the process of annotating an input audio with information that attributes temporal regions of the audio signal to their respective sources, which may include both speech and non-speech events. For speech regions, the diarization system also specifies the locations of speaker boundaries and assign relative speaker labels to each homogeneous segment of speech. In short, speaker diarization systems effectively answer the question of ‘who spoke when’. There are several important applications for speaker diarization technology, such as facilitating speaker indexing systems to allow users to directly access the relevant segments of interest within a given audio, and assisting with other downstream processes such as summarizing and parsing. When combined with automatic speech recognition (ASR) systems, the metadata extracted from a speaker diarization system can provide complementary information for ASR transcripts including the location of speaker turns and relative speaker segment labels, making the transcripts more readable. Speaker diarization output can also be used to localize the instances of specific speakers to pool data for model adaptation, which in turn boosts transcription accuracies. Speaker diarization therefore plays an important role as a preliminary step in automatic transcription of audio data. The aim of this work is to improve the usefulness and practicality of speaker diarization technology, through the reduction of diarization error rates. In particular, this research is focused on the segmentation and clustering stages within a diarization system. Although particular emphasis is placed on the broadcast news audio domain and systems developed throughout this work are also trained and tested on broadcast news data, the techniques proposed in this dissertation are also applicable to other domains including telephone conversations and meetings audio. Three main research themes were pursued: heuristic rules for speaker segmentation, modelling uncertainty in speaker model estimates, and modelling uncertainty in eigenvoice speaker modelling. The use of heuristic approaches for the speaker segmentation task was first investigated, with emphasis placed on minimizing missed boundary detections. A set of heuristic rules was proposed, to govern the detection and heuristic selection of candidate speaker segment boundaries. A second pass, using the same heuristic algorithm with a smaller window, was also proposed with the aim of improving detection of boundaries around short speaker segments. Compared to single threshold based methods, the proposed heuristic approach was shown to provide improved segmentation performance, leading to a reduction in the overall diarization error rate. Methods to model the uncertainty in speaker model estimates were developed, to address the difficulties associated with making segmentation and clustering decisions with limited data in the speaker segments. The Bayes factor, derived specifically for multivariate Gaussian speaker modelling, was introduced to account for the uncertainty of the speaker model estimates. The use of the Bayes factor also enabled the incorporation of prior information regarding the audio to aid segmentation and clustering decisions. The idea of modelling uncertainty in speaker model estimates was also extended to the eigenvoice speaker modelling framework for the speaker clustering task. Building on the application of Bayesian approaches to the speaker diarization problem, the proposed approach takes into account the uncertainty associated with the explicit estimation of the speaker factors. The proposed decision criteria, based on Bayesian theory, was shown to generally outperform their non- Bayesian counterparts.

Re-framing fashion : from original and copy to adaptation

Partington notes that clothing produced by individual consumers through adaptation of patterns is contextualised as a watered down version of original couture. In its most reductive form, this notion characterises fashion as commercial and exploitative. Descriptors such as appropriation, imitation, copy and so forth have restricted the opportunity to understand fashion as a major global cultural form and institution. Therefore exploring and understanding the concept of adaptation will shift the attention from a superficial assessment of original versus imitation or copy to adaptation as a practice that provides a better framework for the understanding of designers’ and couturiers’ innovative practices and creativity, describing also the active engagement of consumers with fashion at the micro level. Adaptation can also provide a way to understand different historical shifts in the fashion system, from individual creative agency with home dressmaking and re-making to the explosion of the mass market and the consequent abandonment of such practices. Home dressmaking has been replaced by fashion remix of mass produced garments, a practice that thrives in our environment of globalised fast fashion. Thus this chapter suggests the need for a contextual requalification of concepts such as original, copy, imitation and copyright, and argues that these categories have been played against each other, but they are in fact interdependent. Today, big labels and conglomerates try to control knowledge and innovation through copyright, but, fashion escapes copyright because, in fashion, creativity is contextual. The institutionalisation of couture from 1868 served as a way to control knowledge about production processes in fashion; on the other hand, adaptation practices, often subversive, have been fundamental to the democratisation of fashion.

The implications of cross-regional differences for the design of In-vehicle Information Systems : a comparison of Australian and Chinese drivers

The increasing global distribution of automobiles necessitates that the design of In-vehicle Information Systems (IVIS) is appropriate for the regions to which they are being exported. Differences between regions such as culture, environment and traffic context can influence the needs, usability and acceptance of IVIS. This paper describes two studies aimed at identifying regional differences in IVIS design needs and preferences across drivers from Australia and China to determine the impact of any differences on IVIS design. Using a questionnaire and interaction clinics, the influence of cultural values and driving patterns on drivers' preferences for, and comprehension of, surface- and interaction-level aspects of IVIS interfaces was explored. Similarities and differences were found between the two regional groups in terms of preferences for IVIS input control types and labels and in the comprehension of IVIS functions. Specifically, Chinese drivers preferred symbols and Chinese characters over English words and were less successful (compared to Australians) at comprehending English abbreviations, particularly for complex IVIS functions. Implications in terms of the current trend to introduce Western-styled interfaces into other regions with little or no adaptation are discussed.

The knock-on from the knock-off : recent shifts within Australian mass market fashion design practice

Australia's mass market fashion labels have traditionally benefitted from their peripheral location to the world's fashion centres. Operating a season behind, Australian mass market designers and buyers were well-placed to watch trends play out overseas before testing them in the Australian marketplace. For this reason, often a designer's role was to source and oversee the manufacture of 'knock-offs', or close copies of northern hemisphere mass market garments. Both Weller and Walsh have commented on this practice.12 The knock-on effect from this continues to be a cautious, derivative fashion sensibility within Australian mass market fashion design, where any new trend or product is first tested and proved overseas months earlier. However, there is evidence that this is changing. The rapid online dissemination of global fashion trends, coupled with the Australian consumer’s willingness to shop online, has meant that the ‘knock-off’ is less viable. For this reason, a number of mass market companies are moving away from the practice of direct sourcing and are developing product in-house under a northern hemisphere model. This shift is also witnessed in the trend for mass market companies to develop collections in partnership with independent Australian designers. This paper explores the current and potential effects of these shifts within Australian mass market design practice, and discusses how they may impact on both consumers and on the wider culture of Australian fashion.

The contribution of psychosocial factors to socioeconomic differences in food purchasing

In developed countries the relationship between socioeconomic position (SEP) and health is unequivocal. Those who are socioeconomically disadvantaged are known to experience higher morbidity and mortality from a range of chronic diet-related conditions compared to those of higher SEP. Socioeconomic inequalities in diet are well established. Compared to their more advantaged counterparts, those of low SEP are consistently found to consume diets less consistent with dietary guidelines (i.e. higher in fat, salt and sugar and lower in fibre, fruit and vegetables). Although the reasons for dietary inequalities remain unclear, understanding how such differences arise is important for the development of strategies to reduce health inequalities. Both environmental (e.g. proximity of supermarkets, price, and availability of foods) and psychosocial (e.g. taste preference, nutrition knowledge) influences are proposed to account for inequalities in food choices. Although in the United States (US), United Kingdom (UK), and parts of Australia, environmental factors are associated with socioeconomic differences in food choices, these factors do not completely account for the observed inequalities. Internationally, this context has prompted calls for further exploration of the role of psychological and social factors in relation to inequalities in food choices. It is this task that forms the primary goal of this PhD research. In the small body of research examining the contribution of psychosocial factors to inequalities in food choices, studies have focussed on food cost concerns, nutrition knowledge or health concerns. These factors are generally found to be influential. However, since a range of psychosocial factors are known determinants of food choices in the general population, it is likely that a range of factors also contribute to inequalities in food choices. Identification of additional psychosocial factors of relevance to inequalities in food choices would provide new opportunities for health promotion, including the adaption of existing strategies. The methodological features of previous research have also hindered the advancement of knowledge in this area and a lack of qualitative studies has resulted in a dearth of descriptive information on this topic. This PhD investigation extends previous research by assessing a range of psychosocial factors in relation to inequalities in food choices using both quantitative and qualitative techniques. Secondary data analyses were undertaken using data obtained from two Brisbane-based studies, the Brisbane Food Study (N=1003, conducted in 2000), and the Sixty Families Study (N=60, conducted in 1998). Both studies involved main household food purchasers completing an interviewer-administered survey within their own home. Data pertaining to food-purchasing, and psychosocial, socioeconomic and demographic characteristics were collected in each study. The mutual goals of both the qualitative and quantitative phases of this investigation were to assess socioeconomic differences in food purchasing and to identify psychosocial factors relevant to any observed differences. The quantitative methods then additionally considered whether the associations examined differed according to the socioeconomic indicator used (i.e. income or education). The qualitative analyses made a unique contribution to this project by generating detailed descriptions of socioeconomic differences in psychosocial factors. Those with lower levels of income and education were found to make food purchasing choices less consistent with dietary guidelines compared to those of high SEP. The psychosocial factors identified as relevant to food-purchasing inequalities were: taste preferences, health concerns, health beliefs, nutrition knowledge, nutrition concerns, weight concerns, nutrition label use, and several other values and beliefs unique to particular socioeconomic groups. Factors more tenuously or inconsistently related to socioeconomic differences in food purchasing were cost concerns, and perceived adequacy of the family diet. Evidence was displayed in both the quantitative and qualitative analyses to suggest that psychosocial factors contribute to inequalities in food purchasing in a collective manner. The quantitative analyses revealed that considerable overlap in the socioeconomic variation in food purchasing was accounted for by key psychosocial factors of importance, including taste preference, nutrition concerns, nutrition knowledge, and health concerns. Consistent with these findings, the qualitative transcripts demonstrated the interplay between such influential psychosocial factors in determining food-purchasing choices. The qualitative analyses found socioeconomic differences in the prioritisation of psychosocial factors in relation to food choices. This is suggestive of complex cultural factors that distinguish advantaged and disadvantaged groups and result in socioeconomically distinct schemas related to health and food choices. Compared to those of high SEP, those of lower SEP were less likely to indicate that health concerns, nutrition concerns, or food labels influenced food choices, and exhibited lower levels of nutrition knowledge. In the absence of health or nutrition-related concerns, taste preferences tended to dominate the food purchasing choices of those of low SEP. Overall, while cost concerns did not appear to be a main determinant of socioeconomic differences in food purchasing, this factor had a dominant influence on the food choices of some of the most disadvantaged respondents included in this research. The findings of this study have several implications for health promotion. The integrated operation of psychosocial factors on food purchasing inequalities indicates that multiple psychosocial factors may be appropriate to target in health promotion. It also seems possible that the inter-relatedness of psychosocial factors would allow health promotion targeting a single psychosocial factor to have a flow-on affect in terms of altering other influential psychosocial factors. This research also suggests that current mass marketing approaches to health promotion may not be effective across all socioeconomic groups due to differences in the priorities and main factors of influence in food purchasing decisions across groups. In addition to the practical recommendations for health promotion, this investigation, through the critique of previous research, and through the substantive study findings, has highlighted important methodological considerations for future research. Of particular note are the recommendations pertaining to the selection of socioeconomic indicators, measurement of relevant constructs, consideration of confounders, and development of an analytical approach. Addressing inequalities in health has been noted as a main objective by many health authorities and governments internationally. It is envisaged that the substantive and methodological findings of this thesis will make a useful contribution towards this important goal.

Segmentation of cone-beam CT using a hidden Markov random field with informative priors

Purpose: Flat-detector, cone-beam computed tomography (CBCT) has enormous potential to improve the accuracy of treatment delivery in image-guided radiotherapy (IGRT). To assist radiotherapists in interpreting these images, we use a Bayesian statistical model to label each voxel according to its tissue type. Methods: The rich sources of prior information in IGRT are incorporated into a hidden Markov random field (MRF) model of the 3D image lattice. Tissue densities in the reference CT scan are estimated using inverse regression and then rescaled to approximate the corresponding CBCT intensity values. The treatment planning contours are combined with published studies of physiological variability to produce a spatial prior distribution for changes in the size, shape and position of the tumour volume and organs at risk (OAR). The voxel labels are estimated using the iterated conditional modes (ICM) algorithm. Results: The accuracy of the method has been evaluated using 27 CBCT scans of an electron density phantom (CIRS, Inc. model 062). The mean voxel-wise misclassification rate was 6.2%, with Dice similarity coefficient of 0.73 for liver, muscle, breast and adipose tissue. Conclusions: By incorporating prior information, we are able to successfully segment CBCT images. This could be a viable approach for automated, online image analysis in radiotherapy.

Segmentation of cone-beam CT using a hidden Markov random field with informative priors

Cone-beam computed tomography (CBCT) has enormous potential to improve the accuracy of treatment delivery in image-guided radiotherapy (IGRT). To assist radiotherapists in interpreting these images, we use a Bayesian statistical model to label each voxel according to its tissue type. The rich sources of prior information in IGRT are incorporated into a hidden Markov random field model of the 3D image lattice. Tissue densities in the reference CT scan are estimated using inverse regression and then rescaled to approximate the corresponding CBCT intensity values. The treatment planning contours are combined with published studies of physiological variability to produce a spatial prior distribution for changes in the size, shape and position of the tumour volume and organs at risk. The voxel labels are estimated using iterated conditional modes. The accuracy of the method has been evaluated using 27 CBCT scans of an electron density phantom. The mean voxel-wise misclassification rate was 6.2\%, with Dice similarity coefficient of 0.73 for liver, muscle, breast and adipose tissue. By incorporating prior information, we are able to successfully segment CBCT images. This could be a viable approach for automated, online image analysis in radiotherapy.

Intrusion detection in distributed systems, an approach based on taint marking

This paper presents a new framework for distributed intrusion detection based on taint marking. Our system tracks information flows between applications of multiple hosts gathered in groups (i.e., sets of hosts sharing the same distributed information flow policy) by attaching taint labels to system objects such as files, sockets, Inter Process Communication (IPC) abstractions, and memory mappings. Labels are carried over the network by tainting network packets. A distributed information flow policy is defined for each group at the host level by labeling information and defining how users and applications can legally access, alter or transfer information towards other trusted or untrusted hosts. As opposed to existing approaches, where information is most often represented by two security levels (low/high, public/private, etc.), our model identifies each piece of information within a distributed system, and defines their legal interaction in a fine-grained manner. Hosts store and exchange security labels in a peer to peer fashion, and there is no central monitor. Our IDS is implemented in the Linux kernel as a Linux Security Module (LSM) and runs standard software on commodity hardware with no required modification. The only trusted code is our modified operating system kernel. We finally present a scenario of intrusion in a web service running on multiple hosts, and show how our distributed IDS is able to report security violations at each host level.

Real-time video event detection in crowded scenes using MPEG derived features : a multiple instance learning approach

This paper presents an investigation into event detection in crowded scenes, where the event of interest co-occurs with other activities and only binary labels at the clip level are available. The proposed approach incorporates a fast feature descriptor from the MPEG domain, and a novel multiple instance learning (MIL) algorithm using sparse approximation and random sensing. MPEG motion vectors are used to build particle trajectories that represent the motion of objects in uniform video clips, and the MPEG DCT coefficients are used to compute a foreground map to remove background particles. Trajectories are transformed into the Fourier domain, and the Fourier representations are quantized into visual words using the K-Means algorithm. The proposed MIL algorithm models the scene as a linear combination of independent events, where each event is a distribution of visual words. Experimental results show that the proposed approaches achieve promising results for event detection compared to the state-of-the-art.

Large-scale analysis of formations in soccer

Due to the demand for better and deeper analysis in sports, organizations (both professional teams and broadcasters) are looking to use spatiotemporal data in the form of player tracking information to obtain an advantage over their competitors. However, due to the large volume of data, its unstructured nature, and lack of associated team activity labels (e.g. strategic/tactical), effective and efficient strategies to deal with such data have yet to be deployed. A bottleneck restricting such solutions is the lack of a suitable representation (i.e. ordering of players) which is immune to the potentially infinite number of possible permutations of player orderings, in addition to the high dimensionality of temporal signal (e.g. a game of soccer last for 90 mins). Leveraging a recent method which utilizes a "role-representation", as well as a feature reduction strategy that uses a spatiotemporal bilinear basis model to form a compact spatiotemporal representation. Using this representation, we find the most likely formation patterns of a team associated with match events across nearly 14 hours of continuous player and ball tracking data in soccer. Additionally, we show that we can accurately segment a match into distinct game phases and detect highlights. (i.e. shots, corners, free-kicks, etc) completely automatically using a decision-tree formulation.