Balancing Protection and Quality in Big Data Analytics Pipelines.

Existing data engine implementations do not properly manage the conflict between the need of protecting and sharing data, which is hampering the spread of big data applications and limiting their impact. These two requirements have often been studied and defined independently, leading to a conceptual and technological misalignment. This article presents the architecture and technical implementation of a data engine addressing this conflict by integrating a new governance solution based on access control within a big data analytics pipeline. Our data engine enriches traditional components for data governance with an access control system that enforces access to data in a big data environment based on data transformations. Data are then used along the pipeline only after sanitization, protecting sensitive attributes before their usage, in an effort to facilitate the balance between protection and quality. The solution was tested in a real-world smart city scenario using the data of the Oslo city transportation system. Specifically, we compared the different predictive models trained with the data views obtained by applying the secure transformations required by different user roles to the same data set. The results show that the predictive models, built on data manipulated according to access control policies, are still effective.

Counseling (ro)bot as a use case for 5G/6G.

This paper presents a counseling (ro)bot called Visual Counseling Agent (VICA) which focuses on remote mental healthcare. It is an agent system leveraging artificial intelligence (AI) to aid mentally distressed persons through speech conversation. The system terminals are connected to servers by the Internet exploiting Cloud-nativeness, so that anyone who has any type of terminal can use it from anywhere. Despite a promising voice communication interface, VICA shows limitations in conversation continuity on conventional 4G networks. Concretely, the use of the current 4G networks produces word dropping, delayed response, and the occasional connection failure. The objective of this paper is to mitigate these issues by leveraging a 5G/6G slice inclusive of mobile/multiple edge computing (MEC). First, we propose and partly implement the enhanced and advanced version of VICA. Servers of enhanced versions collaborate to increase speech recognition reliability. Although it significantly increases generated data volume, the advanced version enables a recognition of the facial expressions to greatly enhance counseling quality. Then, we propose a quality assurance mechanism using multiple levels of catalog, as well as 5G/6G slice inclusive of MEC, and conduct experiments to uncover issues related to the 4G. Results indicate that the number of speech recognition errors in Internet Cloud is more than twofold compared to edge computing, implying that quality assurance using 5G/6G in conjunction with VICA Counseling (ro)bot has higher efficiency.

A Modelling Framework for Evidence-Based Public Health Policy Making.

It is widely recognised that the process of public health policy making (i.e., the analysis, action plan design, execution, monitoring and evaluation of public health policies) should be evidenced based, and supported by data analytics and decision-making tools tailored to it. This is because the management of health conditions and their consequences at a public health policy making level can benefit from such type of analysis of heterogeneous data, including health care devices usage, physiological, cognitive, clinical and medication, personal, behavioural, lifestyle data, occupational and environmental data. In this paper we present a novel approach to public health policy making in a form of an ontology, and an integrated platform for realising this approach. Our solution is model-driven and makes use of big data analytics technology. More specifically, it is based on public health policy decision making (PHPDM) models that steer the public health policy decision making process by defining the data that need to be collected, the ways in which they should be analysed in order to produce the evidence useful for public health policymaking, how this evidence may support or contradict various policy interventions (actions), and the stakeholders involved in the decision-making process. The resulted web-based platform has been implemented using Hadoop, Spark and HBASE, developed in the context of a research programme on public health policy making for the management of hearing loss called EVOTION, funded by the Horizon 2020.

A trusted medical image super-resolution method based on feedback adaptive weighted dense network.

High-resolution (HR) medical images are preferred in clinical diagnoses and subsequent analysis. However, the acquisition of HR medical images is easily affected by hardware devices. As an effective and trusted alternative method, the super-resolution (SR) technology is introduced to improve the image resolution. Compared with traditional SR methods, the deep learning-based SR methods can obtain more clear and trusted HR images. In this paper, we propose a trusted deep convolutional neural network-based SR method named feedback adaptive weighted dense network (FAWDN) for HR medical image reconstruction. Specifically, the proposed FAWDN can transmit the information of the output image to the low-level features by a feedback connection. To explore advanced feature representation and reduce the feature redundancy in dense blocks, an adaptive weighted dense block (AWDB) is introduced to adaptively select the informative features. Experimental results demonstrate that our FAWDN outperforms the state-of-the-art image SR methods and can obtain more clear and trusted medical images than comparative methods.

Bayer Demosaicking With Polynomial Interpolation.

Demosaicking is a digital image process to reconstruct full color digital images from incomplete color samples from an image sensor. It is an unavoidable process for many devices incorporating camera sensor (e.g., mobile phones, tablet, and so on). In this paper, we introduce a new demosaicking algorithm based on polynomial interpolation-based demosaicking. Our method makes three contributions: calculation of error predictors, edge classification based on color differences, and a refinement stage using a weighted sum strategy. Our new predictors are generated on the basis of on the polynomial interpolation, and can be used as a sound alternative to other predictors obtained by bilinear or Laplacian interpolation. In this paper, we show how our predictors can be combined according to the proposed edge classifier. After populating three color channels, a refinement stage is applied to enhance the image quality and reduce demosaicking artifacts. Our experimental results show that the proposed method substantially improves over the existing demosaicking methods in terms of objective performance (CPSNR, S-CIELAB ΔE*, and FSIM), and visual performance.

A rank-ordered marginal filter for deinterlacing.

This paper proposes a new interpolation filter for deinterlacing, which is achievedby enhancing the edge preserving ability of the conventional edge-based line average methods. This filter consists of three steps: pre-processing step, fuzzy metric-based weight assignation step, and rank-ordered marginal filter step. The proposed method is able to interpolate the missing lines without introducing annoying articles. Simulation results show that the images filtered with the proposed algorithm restrain less annoying pixels than the ones acquired by other methods.

Toward sensor-based context aware systems.

This paper proposes a methodology for sensor data interpretation that can combine sensor outputs with contexts represented as sets of annotated business rules. Sensor readings are interpreted to generate events labeled with the appropriate type and level of uncertainty. Then, the appropriate context is selected. Reconciliation of different uncertainty types is achieved by a simple technique that moves uncertainty from events to business rules by generating combs of standard Boolean predicates. Finally, context rules are evaluated together with the events to take a decision. The feasibility of our idea is demonstrated via a case study where a context-reasoning engine has been connected to simulated heartbeat sensors using prerecorded experimental data. We use sensor outputs to identify the proper context of operation of a system and trigger decision-making based on context information.