Multi-detector computed tomography (MDCT) is superior in fracture detection than conventional radiography; however, dose is increased. Cone-beam computed tomography (CBCT) offers higher spatial resolution and lower dose than MDCT. Manufacturers offer an ultra-low-dose algorithm. This study compares the diagnostic accuracy of the ultra-low-dose CBCT (ULDCBCT) with that of the standard-dose CBCT (SDCBCT). In total, 64 patients were scanned with both the SDCBCT and the ULDCBCT protocols. Both studies were reported by two consultant radiologists with fellowship training in emergency radiology separated in time. The reporter recorded a diagnosis of fracture or normal and diagnostic confidence using a 5-point Likert scale. The gold standard was taken as the SDCBCT. Reporters were blinded to the indication and the SDCBCT report. Cases of discrepancy were resolved by consensus. There were 34 fractures and 30 cases had no fracture. Several fractures were missed using the UDCBCT, and there were also several cases of overdiagnosis. ULD was inferior to SD for fracture diagnosis (p < 0.00001). The diagnostic accuracy of ULDCBCT was 82.8% (75.1–88.9 CI). The diagnostic accuracy of plain radiograph was 64% (55.1–75.7% CI). Diagnostic confidence was reduced; the mean confidence for SDCBCT was 4.68 vs 4.12 for ULDCBCT (p < 0.001). The Kappa for interobserver agreement was 0.6. ULDCBCT is inferior to SDCBCT in fracture detection and confidence is reduced. For diagnostic studies, the standard dose should be used.

Automated driving solutions represent a potential solution to promoting driver persistence and the management of fitness to drive issues in older adults. This paper reports on the application of a ‘Human Factors & Ethics Canvas’ and associated methodologies, to support the preliminary specification of an ethically responsible solution for a new driver assistance system. The proposed driving assistance solution has emerged from an analysis of certain ethical principles in relation to the goals and needs of specific older adult drivers (i.e. personae) in different situations (i.e. scenarios). The driving solution is designed to optimize the abilities and participation of older adults.

New assisted driving technology provides a solution to enabling driver persistence while also addressing older adult fitness to drive issues. The proposed driver assistance system follows a detailed literature review, an analysis of secondary data, and the specification of a solution using human machine interaction (HMI) design methods. Overall, the assisted driving concept follows from a principled/ethical perspective in relation to promoting self-efficacy and enablement for older adults. The system is conceptualized as a supportive friend or 'co-pilot'. It is argued that the use of new car-based sensors, along with machine learning intelligence and novel multimodal HMI communication methods will enable driver persistence while also promoting older adult self-efficacy and positive ageing.

Mobility is associated with driving a vehicle.Age-related declines in the abilities of older persons present certain obstacles to safe driving.The negative effects of driving cessation on older adults' physical, mental, cognitive, and social functioning are well reported.Automated driving solutions represent a potential solution to promoting driver persistence and the management of fitness to drive issues in older adults.Technology innovation influences societal values and raises ethical questions.The advancement of new driving solutions raises overarching questions in relation to the values of society and how we design technology (a) to promote positive values around ageing, (b) to enhance ageing experience, (c) to protect human rights, (d) to ensure human benefit and (e) to prioritise human well-being.To this end, this chapter reviews the relevant ethical considerations in relation to assisted driving solutions.Further, it presents a new ethically aligned system concept for assisted driving.It is argued that human benefit, well-being and respect for human identity and rights are important goals for new automated driving technologies.Enabling driver persistence is an issue for all of society and not just older adult.

The optimal locations of transmitters in a building, that best meet the bandwidth requirements of all users, was studied in the past by a small number of authors. They used empirical path loss models to constrain an optimization cost function. This paper concentrates on using a signal-to-interference ratio cost function with realistic field parameters obtained using a fast raytracing algorithm. The solution obtained is shown to perform well in realistic scenarios, meeting a large proportion of user bandwidth requirements. However, it does not perform well in an oversaturated environment.

Many applications, such as autonomous navigation, urban planning, and asset monitoring, rely on the availability of accurate information about objects and their geolocations. In this paper, we propose the automatic detection and computation of the coordinates of recurring stationary objects of interest using street view imagery. Our processing pipeline relies on two fully convolutional neural networks: the first segments objects in the images, while the second estimates their distance from the camera. To geolocate all the detected objects coherently we propose a novel custom Markov random field model to estimate the objects' geolocation. The novelty of the resulting pipeline is the combined use of monocular depth estimation and triangulation to enable automatic mapping of complex scenes with the simultaneous presence of multiple, visually similar objects of interest. We validate experimentally the effectiveness of our approach on two object classes: traffic lights and telegraph poles. The experiments report high object recall rates and position precision of approximately 2 m, which is approaching the precision of single-frequency GPS receivers.

This paper describes a point-to-multipoint 3-D convex space-based ray-tracing technique. This visibility list is calculated and stored and can be reused as needed. What distinguishes our method is that the visibility list is transmitter location-independent, is a 3-D implementation, and is highly computationally efficient. The division of the building into free and filled convex spaces leads to an efficient Method of Images reflection and diffraction path generation algorithm. This technique can be used to optimize the locations of base transceivers in a highly efficient manner. The first step in producing this tool is the generation of efficient ray-tracing algorithms. The ray-tracing algorithm was specifically designed for later incorporation into a transmitter optimization algorithm. This requires a fast ray-tracing method because of its computationally intensive needs-running multiple times over a point-to-multipoint grid. Our algorithm is executed for sample building environments and then for a real building and compared with measurements to confirm its validity. It is clear that the results are in good agreement but do indicate that a highly accurate spatial modeling of the building is required.

We describe the motivation behind, design, and implementation of MILLA, a prototype speech-to-speech English language tutoring system. 1 Background Learning a new language involves the acquisition and integration of a range of skills. A human tutor aids learners by (i) providing a framework of tasks suitable to the learner’s needs, (ii) monitoring learner progress and adapting task content and delivery style to suit, and (iii) providing a source of speaking practice and motivation. With the advent of audiovisual technology and the communicative paradigm in language pedagogy, focus has shifted from written grammar and translation to activities focusing more on communicative competence in listening and spoken production. In recent years the Common European Framework of Reference for Language Learning and Teaching (CEFR) officially added a more integrative fifth skill – spoken interaction to the traditional four skills – reading and listening, and writing and speaking (Little, 2006) . While second languages have always been learned interactively through negotiation of meaning between speakers of different languages sharing living or working environments, these methods did not figure in formal (funded) settings. However, with increased mobility and globalisation, many formal learners now need language as a practical tool for everyday life and business rather than simply as an academic achievement. Developments in Computer Assisted Language Learning (CALL) have resulted in a range of free and commercial language learning material for autonomous study. Much of this material transfers long-established paper-based and audiovisual exercises to the computer screen. Pronunciation training exercises have been developed which provide feedback either through the learner listening back to their own efforts and comparing to a model, or the system providing a score or other feedback. These resources are very useful in development of discrete skills, but the challenge of providing spoken interaction tuition and practice remains. The MILLA system, developed at the 2014 eNTERFACE workshop (‘The 10th International Summer Workshop on Multimodal Interfaces eNTERFACE’14 ISCA Training School’, 2014) is a multimodal spoken dialogue system combining custom modules with existing web resources in a balanced curriculum, and, by integrating spoken dialogue, modelling some of the advantages of a human tutor. 2 MILLA System Components MILLA’s spoken dialogue Tuition Manager consults a two-level curriculum of language learning tasks, a learner record, and a learner state module to greet and enroll learners, direct them to language learning submodules, provide feedback and scoring, and monitor user state with Kinect sensors. All of the tuition manager’s interaction with the user can be performed using speech through a Cereproc TTS voice using Cereproc’s Python SDK (‘CereVoice Engine Text-to-Speech SDK | CereProc Text-to-Speech’, 2014) and understanding via CMU’s Sphinx4 ASR (Walker et al., 2004) through custom Python bindings using W3C compliant Java Speech Format Grammars. Tasks include spoken dialogue practice with two different chatbots, first language (L1) focused and general pronunciation training, and grammar and vocabulary exercises. Several speech recognition (ASR) engines (HTK, Google Speech) and text-to speech (TTS) voices (Mac and Windows system voices, Google Speech) are incorporated in the modules to meet the demands of particular tasks and to provide a cast of voice characters which provide a variety of speech models to the learner. Microsoft’s Kinect SDK (‘Kinect for Windows SDK’, 2014) is used for gesture recognition and as a platform for affect recognition. The tuition manager and all interfaces are written in Python 2.6, with additional C#, Javascript, Java, and Bash coding in the Kinect, chat, Sphinx4, and pronunciation elements. For rapid prototyping many of the dialogue modules were first written in VoiceXML, and then ported to Python modules. 2.1 Pronunciation Tuition MILLA incorporates two pronunciation modules, based on comparison of learner production with model production: (i) a focused pronunciation tutor using HTK ASR with the five-state 32 Gaussian mixture monophone acoustic models provided with the Penn Aligner toolkit (Young, n.d.; Yuan & Liberman, 2008) on the system’s local machine and (ii) MySpeech a phrase level trainer hosted on University College Dublin’s cluster and accessed by the system via Internet (Cabral et al., 2012). For the focused pronunciation system, we used the baseline implementation of the Goodness of Pronunciation algorithm, (Witt & Young, 2000). GOP scoring involves two phases: 1) a free phone loop recognition phase which determines the most likely phone sequence given the input speech without giving the ASR any information about the target sentence, and 2) a forced alignment phase which provides the ASR system with the orthographic transcription of the input sentence and force aligns the speech signal with the expected phone sequence. For each phone realization aligned to the speech signal, comparison of the log-likelihoods of the forced alignment and thevfree recognition phases, produces a GOP score where zero reflects a perfect match and increasing positive scores correspond to inaccuracies. Phone specific threshold scores were set to decide whether a phone was mispronounced (``rejected'') or not (``accepted''), by artificially inserting errors in the pronunciation lexicon and running the algorithm on native recordings, as in (Kanters, Cucchiarini, & Strik, 2009). After preliminary testing, we constrained the free phone loop recogniser for more robust behavior, using phone confusions common in specific L1’s to define constrained phone grammars. A database of common errors in several L1s with test utterances was built into the curriculum module. 2.2 Spoken Interaction Tuition (Chat) To provide spoken interaction practice, MILLA sends the user to Michael (Level1) or Susan (Level 2), two chatbots created using the Pandorabots web-based chatbot hosting service . The bots were first implemented in text-to-text form in AIML (Artificial Intelligence Markup Language) and then TTS and ASR were added through the Web Speech API, conforming to W3C standards (W3C, 2014). Based on consultation with language teachers and learners, the system allows users to speak directly to the chat bot, or enter chat responses using text input. A chat log was also implemented into the interface, allowing the user to read back or replay several of their previous interactions with the chat bot. 2.3 Grammar, Vocabulary and External Resources MILLA’s curriculum includes a number of graded activities from the OUP’s English File and the British Council’s Learn English websites (REF). Wherever possible the system scrapes any scores returned for exercises and incorporates them into the learner’s record, while in other cases the progression and scoring system includes a time required to be spent on the exercises before the user progresses to the next exercises. There are also a number of custom morphology and syntax exercises designed for MILLA using Voxeo’s Prophecy platform and VoiceXML which will be ported to MILLA in the near future. 2.4 User State and Gesture Recognition MILLA includes a learner state module which will eventually infer boredom or involvement in the learner. As a first pass, gestures indicating various commands were designed and incorporated into the system using Microsoft’s Kinect SDK. The current implementation comprises four gestures (Stop, I don’t know, Swipe Left/Right), which were designed by tracking the skeletal movements involved and extracting joint coordinates on the x,y, and z planes to train the recognition process. As MILLA is multiplatform (Unix and Windows), Python’s socket programming modules was used to communicate between the Windows machine running the Kinect and the Mac laptop hosting MILLA.

The purpose of this work is to provide a fast multipoint ray-tracing algorithm when the location of many transmitters is not fixed. When optimising the location of multiple transmitters numerically using an optimization algorithm the visibility algorithm for reflective and diffractive surfaces is normally computed each time the transmitter is moved. However when splitting the building into a set of convex spaces, the visibility algorithm does not need to be changed at any iteration of the optimisation process, since the convex spaces inherently provide the visibility algorithm themselves.

Heliophysics is the branch of physics that investigates the interactions and cor- relation of dierent events across the Solar System. The mathematical models that describe and predict how physical events move across the solar system (ie. Propagation Models) are of great relevance. These models depend on parame- ters that users must set, hence the ability to correctly set the values is key to reliable simulations. Traditionally, parameter values can be inferred from data either at the source (the Sun) or arrival point (the target) or can be extrapo- lated from common knowledge of the event under investigation. Another way of setting parameters for Propagation Models is proposed here: instead of guess- ing a priori parameters from scientific data or common knowledge, the model is executed as a parameter-sweep job and selects a posteriori the parameters that yield results most compatible with the event data. In either case (a priori and a posteriori), the correct use of Propagation Models requires information to either select the parameters, validate the results, or both. In order to do so, it is necessary to access sources of information. For this task, the HELIO project proves very eective as it oers the most comprehensive integrated information system in this domain and provides access and coordination to services to mine and analyze data. HELIO also provides a Propagation Model called SHEBA, the extension of which is currently being developed within the SCI-BUS project (a coordinated eort for the development of a framework capable of oering to science gateways seamless access to major computing and data infrastructures).

Heliophysics is the branch of physics that investigates the interactions and cor-relation of different events across the Solar System. The mathematical modelsthat describe and predict how physical events move across the solar system (ie.Propagation Models) are of great relevance. These models depend on parame-ters that users must set, hence the ability to correctly set the values is key toreliable simulations. Traditionally, parameter values can be inferred from dataeither at the source (the Sun) or arrival point (the target) or can be extrapo-lated from common knowledge of the event under investigation. Another way ofsetting parameters for Propagation Models is proposed here: instead of guess-ing a priori parameters from scientific data or common knowledge, the model isexecuted as a parameter-sweep job and selects a posteriori the parameters thatyield results most compatible with the event data. In either case (a priori anda posteriori), the correct use of Propagation Models requires information toeither select the parameters, validate the results, or both. In order to do so, itis necessary to access sources of information. For this task, the HELIO projectproves very effective as it offers the most comprehensive integrated informationsystem in this domain and provides access and coordination to services to mineand analyze data. HELIO also provides a Propagation Model called SHEBA,the extension of which is currently being developed within the SCI-BUS project(a coordinated effort for the development of a framework capable of offering toscience gateways seamless access to major computing and data infrastructures).

If several distributed and disparate computer resources exist, many of whichhave been created for different and diverse reasons, and several large scale com-puting challenges also exist with similar diversity in their backgrounds, then oneproblem which arises in trying to assemble enough of these resources to addresssuch challenges is the need to align and accommodate the different motivationsand objectives which may lie behind the existence of both the resources andthe challenges. Software agents are offered as a mainstream technology formodelling the types of collaborations and relationships needed to do this. Asan initial step towards forming such relationships, agents need a mechanism toconsider social and economic backgrounds. This paper explores addressing so-cial and economic differences using a combination of textual descriptions knownas social profiles and search engine technology, both of which are integrated intoan agent technology.

An important aspect of EMI is the delivery of 'quality software'.For this reason the quality assurance (QA) group was introduced.There are a key number of beneficiaries of this work in several work packages.These EMI development and support activities are required to produce key performance indicators (KPIs) and metrics for milestone, quarterly and yearly deliverables based on the information provided by the QA group.However, EMI has a large number of varying sized products, different and existing middlewares and various bug/feature request for change (RfC) trackers used by each product team.The only way to reliably produce KPIs and metrics related to change management in such a varied project is to introduce simplifying, common environments that are readily accessible by all the different customers of the project.For this reason an extensible XML-based framework was defined for storing, plotting, querying and tabulating change tracker information for all its customers.

HELIO [8] is a project funded under the FP7 program for the discovery and analysis of data for heliophysics. During its development, standards and common frameworks were adopted in three main areas of the project: query services, processing services, and the security infrastructure. After a first, proprietary implementation of the security service, it was suggested moving it to a standard security framework to simplify the enforcement of security on the different sites. As the HELIO front end is built with Spring and the TAVERNA server (HELIO workflow engine) has a security framework compatible with Spring, it has been decided to move the CIS in Spring security [2]. HELIO has two different processing services: one is a generic processing service called HELIO Processing Services (HPS), the other is called Context Service (CTX) and it runs specific IDL procedures. The CTX implements the UWS [4] interface from the IVOA [5], a standard interface for job submission used in the helio and astro-physics community. In its final release, the HPS will expose an UWS compliant interface. Finally, some of the HELIO services perform queries, to simplify the implementation and usage of this services a single query interface (the HELIO Query Interface) has been designed for all these services. The use of these solutions for security, execution, and query allows for easier implementation of the original HELIO architecture and for a simpler deployment of the services.

The EMI Quality Model has been created to define, and later review, the EMI (European Middleware Initiative) software product and process quality. A quality model is based on a set of software quality metrics and helps to set clear and measurable quality goals for software products and processes. The EMI Quality Model follows the ISO/IEC 9126 Software Engineering Product Quality to identify a set of characteristics that need to be present in the EMI software. For each software characteristic, such as portability, maintainability, compliance, etc, a set of associated metrics and KPIs (Key Performance Indicators) are identified. This article presents how the EMI Quality Model and the EMI Metrics have been defined in the context of the software quality assurance activities carried out in EMI. It also describes the measurement plan and presents some of the metrics reports that have been produced for the EMI releases and updates. It also covers which tools and techniques can be used by any software project to extract "code metrics" on the status of the software products and "process metrics" related to the quality of the development and support process such as reaction time to critical bugs, requirements tracking and delays in product releases.

The European Middleware Initiative (EMI) is the collaboration of the major European middleware providers, ARC, gLite, UNICORE, and dCache. It aims to deliver a consolidated set of middleware components for deployment in EGI and PRACE, extend the interoperability and integration between grids and other computing infrastructures, strengthen the reliability and manageability of the services and establish a sustainable model to support, harmonise and evolve the middleware, ensuring it responds to the requirements of the scientific communities relying on it. EMI will carry out the collective task of supporting and maintaining the middleware for their user communities. In order to enable the infrastructures to achieve this task, the middleware services must play an important role and mark a clear transition to more sustainable models by adopting best-practice service provision methods such as the ITIL processes or the ISO guidelines for software quality and validation. Repositories of packages, status reports, quality metrics and test and compliance programs are created and maintained to support the project software engineering activities and other providers of applications and services based on the EMI middleware. This article reports on the initial work of the EMI project and the solutions adopted for the software releases, development processes, quality compliance metrics and distribution repositories.