Previous Departmental Seminars

2016

Biometrics on Mobile Devices: Opportunities and Challenges

Monday 28 November 2016

Conventional research into biometric systems has focussed on the development of algorithmic solutions to enhance verification and identification of subjects. Typically these solutions have been used in ‘fixed’ implementations, where sensors and systems are deployed in controlled environments, often, during development, in laboratory conditions. Biometric system performance has been shown to drop considerably in public settings due to poor/confusing processes, feedback or instructions given to the user. Minimising environmental effects and interaction problems are critical as biometrics become commonplace on mobile devices – particularly given the wider user-base and usage frequency. Although widely deployed there are a considerable number of research issues that require solutions to ensure accurate performance. The range of sensors within mobile devices raises the possibility of novel implementations for automated identity management, however these opportunities must be balanced by the additional risks and environmental challenges of mobile systems. In this presentation the key opportunities and challenges for research within mobile biometrics will be highlighted, addressing issues such as continuous authentication, multimodality and interaction and vulnerability assessment.

Perspectives on System-level MPSoC Design Space Exploration

Monday 21 November 2016

Embedded systems traditionally have stringent design requirements in terms of (real-time) performance, power consumption, cost and form factor, and system reliability. To provide a good trade-off with respect to these system qualities, embedded systems increasingly use heterogeneous multiprocessor system-on-chip (MPSoC) architectures, deploying components that range from various types of programmable cores to fully dedicated hardware blocks. Designing these complex, heterogeneous systems requires system-level design-space exploration (DSE) technology to support the process of making design decisions. During such DSE, a large variety of different design alternatives can be explored, such as the number and type of processors deployed in the system architecture, the type of interconnection network used to connect system components, or the spatial binding and temporal binding of application tasks to processor cores. Since this process of DSE is highly challenging because the design space that needs to be explored is typically vast, significant research has been performed in the embedded systems community on devising efficient and effective DSE techniques and methods. In this talk, I will provide an overview of the recent advances in our research on system-level MPSoC design space exploration and will present some future challenges that need to be addressed in this domain.

Internet of Things and Data Analytics for Smarter Cities and eHealth 

Monday 7 November 2016

Digital data and the connected worlds of physical objects, people and devices are rapidly changing the way we work and interact with our surrounding environments. New technologies have a profound impact on different domains, such as healthcare, environmental monitoring, urban systems and control and management applications, among others. However, these large volumes of heterogeneous data require new methods of communication, aggregation, processing and mining in order to extract efficient and actionable information from the raw data. This talk will focus on Internet of Things technologies and will discuss the opportunities and challenges involved in using large-scale data analytics in smart city and healthcare applications and services. The talk will also discuss the National Health Services (NHS) funded Internet of Things project for dementia care, called TIHM (https://www.england.nhs.uk/ourwork/innovation/test-beds/tihm/). It is estimated that there will be one million people with dementia in the UK by 2025. Beyond those directly affected, dementia also has a much wider impact on carers, families and health and care services, with social isolation being a key confounding factor. The TIHM project focuses on building the technical and physical infrastructure for an Internet of Things framework to support dementia care and remote monitoring. TIHM uses distributed and secure technologies to capture, analyse and manage the data in order to provide real-time and actionable information to different support groups in dementia care – i.e. clinical, social care, and patients (whenever and wherever appropriate).

4th Engineering Education Research Network Annual Symposium

Thursday 3 November 2016

A two day conference focusing on issues facing engineering education research.

Bio-electronic Medicines – The Pathway to Health

Monday 31 October 2016

The functions and organ systems of our body are, to a significant extent, controlled by the electrical signals that travel along our nerves. Advances in modern microelectronics have paved the way for devices that aim to control these biological processes and treat disease by modulation of electrical impulses. Bio-electronic medicines could allow us to address diseases that have so far been untreatable, and target others with greater efficacy, precision and fewer side effects than with conventional molecular medicines. The CAST (Centre for Advanced Sensor Technology) group at the University of Bath aims to develop novel sensor assemblies for both in and ex­-vivo studies of a wide range of different bio-electronic applications. One project is directly involved with recording neural signals from the peripheral nervous system using velocity selective recording, and aims to significantly improve the state of the art for implantable recording systems and closed loop devices. A second project is centred on the development of electrical biosensors for roles from drug discovery to the detection of cancer biomarkers. This work focusses on the simple modification of conventional CMOS devices for cheap and readily available clinical applications.

Analysis of Information Encoding and Dynamics in Optically Recorded Cortical Circuits

Monday 24 October 2016

In recent years, technology for large-scale recording of neural circuit dynamics, at single cell resolution, has progressed extremely rapidly. Several international initiatives, including the US NIH BRAIN Initiative, mean that we are likely to see further developments, including the ability to manipulate as well as read out neural ensemble activity. As well as enhancing our understanding of numerous basic questions in systems neuroscience, we can hope that these techniques are likely to be of translational benefit, by allowing the characterisation of changes to circuit behaviour in mouse models of neurodegenerative disorders to be studied in great detail and across scales. Scalable data analysis tools capable of taking into consideration patterns of neural ensemble activity, however, become a limiting factor once neural population sizes exceed a few tens of neurons. In the past, I have developed information theoretic methods for analysing how information is represented in spike trains fired by small ensembles of neurons. In this talk, I will describe several approaches we are taking to scale these approaches up to tens and hundreds of neurons recorded simultaneously through two photon calcium imaging. We take two quite different approaches. In the first approach, we consider the calcium time series from the neural ensemble as a multivariate continuous time series, and employ approaches from nonlinear dynamics, together with dimensionality reduction. In the second, we use a calcium transient detection algorithm to instead represent the data as a digitized multineuron spike train, and make strong but testable assumptions about the underlying variability. In the talk, I will describe the application of these methods to data from a number of cortical circuits: neocortical, archicortical (hippocampus) and the cerebellar cortical circuit.

Electronics and Technology in Agriculture - the Next Farming Revolution?

Monday 17 October 2016

Advances in technology, especially electronics, is driving a new revolution in agriculture and horticulure, partly as a response to global challenges such as food shortage, climate changes and increasing population. This is evident by the various terms appearing in the literature and press including "E-agriculture", "Precision Farming", "Agritech", "Connected Farms", "Internet of Agri-things" and even "the Internet of Tomatoes". The seminar aims to show how electronics is at the heart of technological advances and will cover a range of topics including automatic milking machines, robot weed sprayers, GPS, driverless tractors, automated nitrogen sensing for crop spraying, wireless sensor networks, IoT, satellite and airborne remote sensing, LEDs for crop growth control and pest detection. It will conclude with a look at future develoments such as fully automated crop growing.

Natural Solutions to Electronic Systems Design Problems

Monday 3 October 2016

The increasing versatility, performance, compactness and power efficiency of today’s electronic systems is achieved by pushing technology to its physical limits: systems are increasing in size and complexity comprising thousands of subsystems made of billions of devices, requiring sophisticated programming and control; the devices themselves become smaller and smaller and have reached the atomic scale, which leads to stochastic variations when fabricating them. This makes components more noisy and unreliable and designing reliable systems extremely challenging. In this respect, technological systems are far behind biological organisms which have long since accomplished the feat of not only operating reliably with highly variable components, but also maintaining and tuning themselves in changing environments, when faults occur or they are otherwise perturbed. Biological mechanisms enabling this have co-evolved with the organisms, hence, are perfectly adapted to the requirements of their embodiment. In this context, evolutionary hardware is about hardware that offers the capability to change its structure and behaviour in order to automatically optimise its operation for a specific task or environment, taking inspiration from biological organisms with natural evolution as nature’s guiding optimisation principle. In the talk I will give examples of evolutionary computation applied to electronic design optimisation and performance analysis, and how bio-inspired decision making can improve throughput and fault tolerance of networks on chip.

Time-domain measurement of terahertz plasmons in a semiconductor two-dimensional electron system

Monday 6 June 2016

Terahertz measurement of low-dimensional semiconductor nanostructures can reveal a range of physical phenomena. However, the diffraction limit of free-space terahertz (THz) radiation always impede the probe of elementary excitation in individual nanostructures. Here, we demonstrate a planar terahertz frequency plasmonic circuit which integrate on-chip THz time-domain spectrometer with GaAs/AlGaAs heterostructure which contains a two-dimensional electron system (2DES). THz frequency range plasmons are excited by injecting picosecond-duration pulses, generated and detected by a photoconductive semiconductor, into the 2DES channel. Using voltage modulation of a Schottky gate overlying the 2DES, we observe electrostatic manipulation of 2D plasmon resonances in the 2DES plasmonic cavity. With the change of the magnetic field normal to 2DES plane, the evolution of THz frequency range magnetoplasmon resonance in time domain was measured. The experimental results were found to be in agreement with calculations using a mode-matching approach. Our work introduces a generic technique suitable for studying ultrafast carrier dynamics in low-dimensional semiconductor nanostructures at THz frequencies.

User Centric Backhauling for 5G

Monday 16 May 2016

5G Communication Systems will have challenging requirements for data rates, energy efficiency and the end-to-end latency of the system. To design a system that meets all these challenges several advanced signal processing techniques have been proposed that have implications on the backhaul as well as the fronthaul (for distributed antenna arrays). This talk will focus on available backhaul solutions and their suitability for the 5G challenges. We will also discuss the concept of joint radio access and backhaul optimisation. Some recent results will be shared and discussed.

The Human Connectome: Linking Brain Network Features to Healthy and Pathological Information Processing

Monday 9 May 2016

Our work on connectomics over the last 15 years has shown a small-world, modular, and hub architecture of brain networks [1,2]. Small-world features enable the brain to rapidly integrate and bind information while the modular architecture, present at different hierarchical levels, allows separate processing of various kinds of information (e.g. visual or auditory) while preventing wide-scale spreading of activation [3]. Hub nodes play critical roles in information processing and are involved in many brain diseases [4]. Nonetheless, general observations of human brain connectivity, or of patients at the group-level, have so far had little impact on understanding cognition, or deficiencies in cognition, in individual subjects. As a result, human connectome information is not used as a biomarker for diagnosis or a predictor of the most suitable treatment strategy. After discussing the organisation of brain networks, we will show how connectivity can be used to determine the disease type of individual dementia patients. An important aspect of these brain networks is their spatial organisation in terms of the length of fibre tracts and the location of brain regions [5]. However, simply observing connectivity is insufficient as small changes in network organisation might lead to large changes in network behaviour (dynamics) [6]. We therefore show how simulations can be applied to predict regions that are involved in neural processes. For epilepsy, simulations show us which regions are involved [7], which treatment approach should be used, and whether surgical intervention will be successful or not. We conclude with the role of simulations in understanding the developmental origin of diseases as determining these origins will again inform diagnosis and treatment (http://www.greenbrainproject.org/ ). These are first steps towards using connectome-based computer simulations as a tool to understand normal and pathological processing in individuals. Developing models that are based on anatomical information will be crucial to define the most suitable intervention [8].

Hardware-Software Co-design of Energy-Efficient Many-Core Embedded Systems

Friday 6 May 2016

The functions and organ systems of our body are, to a significant extent, controlled by the electrical signals that travel along our nerves. Advances in modern microelectronics have paved the way for devices that aim to control these biological processes and treat disease by modulation of electrical impulses. Bio-electronic medicines could allow us to address diseases that have so far been untreatable, and target others with greater efficacy, precision and fewer side effects than with conventional molecular medicines. The CAST (Centre for Advanced Sensor Technology) group at the University of Bath aims to develop novel sensor assemblies for both in and ex¬-vivo studies of a wide range of different bio-electronic applications. One project is directly involved with recording neural signals from the peripheral nervous system using velocity selective recording, and aims to significantly improve the state of the art for implantable recording systems and closed loop devices. A second project is centred on the development of electrical biosensors for roles from drug discovery to the detection of cancer biomarkers. This work focusses on the simple modification of conventional CMOS devices for cheap and readily available clinical applications.

Engineering Simulations to Understand and Gain Inspiration From Biological Systems

Monday 25 April 2016

In several areas of biological science, computer simulation is increasingly providing researchers with relevant tools to interrogate human biology, rather than a continued reliance on predictions from animal experiments. Our motivation for simulating biological systems is driven by three key objectives: to further understand biological systems, to further reduce animal experimentation, and inspire development of novel bio-inspired engineering approaches. Yet simulations designed to explore complex biological mechanisms capture a range of uncertain factors that impact the relationship between a derived prediction and the real world: factors that together act as a barrier to wider simulation use and acceptance in further research studies or clinical trials. Developing robust, evidence-based approaches to simulation design, implementation, and analysis has been a feature of research conducted in the York Computational Immunology Lab for a number of years now, within significant immunological case studies. In October I joined the Department of Electronics, with the aim of switching the focus of our research from solely capturing individual biological systems of interest to examining the engineering processes by which biological simulations should be composed, implemented, and applied. In this talk I will give an overview of what we are hoping to achieve over the next four years, how we are hoping to characterise, quantify, and propagate the uncertain factors in simulation development, and how we hope our work will impact bio-inspired engineering approaches both in robotics and further afield.

Implantable Microsystems for Personalised Anti-Cancer Therapy

Monday 18 April 2016

A malignant tumour is an uncontrolled growth of cells that can spread (metastasise) to distant organs. The microenvironment of a tumour is key to this spread and also to the tumour’s resistance to radiotherapy and chemotherapy. For example, well-oxygenated human tumours are 2-3 times more vulnerable to RT than are hypoxic tumours and advanced RT technologies can deliver radiation with sub-mm accuracy to a hypoxic volume. This talk will describe a programme to develop in vivo, wireless biosensors of hypoxia and other key parameters of cancer biology that will allow a patient’s tumour biology to be monitored continuously and without further intervention. The sensors are being integrated on a wireless silicon substrate. The initial system will integrate existing sensing technologies on one miniaturised device (estimated to be 7mm500μm500μm to facilitate minimally-invasive insertion) to probe physical characteristics (temperature, pH, O2) of the microenvironment that give an indirect measure of tumour biology. The talk will also describe novel electrochemical sensors of direct and specific tumour biomarkers.

Calibrated Auralizations and Voice Directivity in Perceptual Room Acoustics Studies

Tuesday 15 March 2016

Room acoustic auralizations are a means by which we can create virtual architectural spaces with coherent acoustical information. Often used for publicity or demonstration purposes, there has been little discussion about the realism and accuracy of auralizations. A method for calibrating room acoustic auralizations has been developed and tested for several existing spaces in Paris, including the Notre Dame Cathedral. Objective measures and subjective listening tests have been carried out to examine the similarity and differences between the actual and simulated room acoustic. In addition to the room acoustic model, the acoustic properties of the source also contribute to the realism of the virtual simulation. A study is currently underway examining the effect of directivity pattern for a vocal source, with the aim of creating a dynamic model which will allow for a more natural source rotation and directivity. Results of preliminary subjective listening tests will be presented.

Sequence Analysis Based Hyperheuristics for Search and Optimisation

Monday 7 March 2016

Online selection hyperheuristics are methods that operate at the level above metaheuristics and adapt themselves to the problem during optimisation. These algorithms use learning techniques to determine the best low-level heuristics (which can include evolutionary operators such as mutation) for a particular problem domain during the optimisation. This talk describes a new online heuristic selection algorithm known as the sequence based selection hyperheuristic (SSHH), based on the principles of sequence analysis used in bioinformatics. The method is able to learn the application probabilities of each of the low level heuristics to the problem individually and as part of unlimited-length sequences, creating new optimisation algorithms on the fly. Results will be presented that show that this method produces state of the art performance across a number of optimisation problem domains including operations research, timetabling and real-world water distribution optimisation problems.

Performance and design of MIMO full-duplex links

Monday 29 February 2016

Multi-antenna full-duplex (FD) technology enables simultaneously reception and transmission of data signals, increasing the spectral efficiency of wireless transmission significantly. The immediate penalty of FD operation is unavoidable loopback self-interference that has to be mitigated by using analog and / or digital cancellation techniques. Due to technical limitations in transceiver electronics, however, perfect mitigation may not be feasible in practice and some residual self-interference remains. In this talk, we review some applications of FD-MIMO and consider the achievable rates of large bidirectional systems under the assumption of residual self-interference due to imperfect mitigation. The analysis is based on the replica method and results from asymptotic random matrix theory.

Three - the journey from worst to first

Monday 15 February 2016

Over the past century, telecommunications technology has changed drastically. Consumer demand for faster and more reliable network has impacted the telecommunications industry across the world; and along with it, a new generation of businesses have grown to join the world market. In the UK, Three launched in 2003 with its unique proposition of 3G in an attempt to combine the internet and mobile together. Despite many challenges, the company has rapidly changed within the past few years from the worst mobile operator to the first. This talk will cover the company’s journey, its commercial strategy, changing public perception, continued opportunities, and my role as a financial reporting analyst.

Microelectronics for sensing, imaging and healthcare

Monday 8 February 2016

Microelectronics technology, and especially CMOS, has revolutionised communications, information and computing technologies over the last 30 years. It has also become clear that the same technology can make excellent image sensors, so changing the face of modern photography. In this talk I will discuss further extensions to the capability and application of CMOS technology, achieved by a combination of new chips design, and heterogeneous integration with optical and biochemical technologies. The talk will be illustrated with examples using nanophotonics for filter technologies on CMOS, metamaterials for terahertz imaging, and ion sensitive field effect transistors for using applications such as chemical monitoring, DNA sequencing, and biochemical assay. These examples will illustrate future directions for the ever expanding range of capabilities of CMOS in the years ahead.

Experiences with edge mining the Internet of Things

Monday 1 February 2016

In this talk, I will look at the concept of edge mining --- data mining that takes place at the edge of a wireless sensor network. Initially, we want our sensor networks to be as flexible as possible but as the area matures, automating information extracting into continuous analytics and pushing such analytics to the edge or leaf nodes of a network can reduce the infrastructural cost of the sensor network and thus enable many applications that would otherwise be infeasible. The talk will be rooted in my experience with deploying wireless sensor nets and deriving meaningful information from the resultant data streams.

Signal Processing for Robot Audition

Monday 25 January 2016

Most current robots have only rudimentary capabilities for audition. The list of desirable audio processing functionality includes the detection of one or more talkers, talker localization and tracking, environment mapping and speech recognition. These processing tasks are much more challenging in real-world environments involving noise, reverberation and conversational spontaneous speech. This talk will start with a brief discussion on the potential roles for signal processing in robot audition and the relationship to computational linguistics. A specific example will then be explored in more detail. The talk will then consider the effects due to combinations of noise with reverberation on speech quality, speech intelligibility and speech recognition performance. Recent results will be given showing the state-of-the-art performance of various combinations of noise reduction, beamforming and dereverberation processing. Conclusions will be drawn so as to identify research threads with strong potential to improve robot audition through the use of signal processing approaches in realistic acoustic scenarios.

How to Use Networks to Manage Networks-on-Chip?

Monday 18 January 2016

In the post-Moore’s-law era, the number of cores on a chip continues to grow as it is still far from reaching the physical and technological limits. This capability provides a solid foundation for advances in high-performance computing and embedded systems, enabling applications such as smart sensors and big data crunching. However, the effective management of many-core systems and networks to support high performance and energy efficiency remains a major challenge. Enormous amounts are spent on improving energy efficiency, without which large swaths of chip area cannot operate at full capacity. This seminar presents a novel approach to manage thousands of cores and their interconnections using a smart and embraced network. The network model capturing a new class of dynamic programming in a network formation will first be described. It supports run-time adaptive optimization for the cores and networks-on-chip (NoC) using a parallel network architecture. It is self-contained and only consumes a few percent of the overall device area and power. The seminar will review a number of advances in this technology, including maximizing on-chip communication performance, optimizing energy efficiency of thousands of cores, and fault recovery and wear leveling. Finally, an extension to cover 3-D architectures and to support thermal adaptation will be described.

Administrative Enquiries:

Please contact Helen Smith, Admissions and Research Student Office, for more information.

Links:

Electronics Department Student Seminars

Chemistry Department Seminars

Computer Science Department Seminars

Physics Department Seminars

 

Events

For past events please see the relevant year: 
201620152014 | 2013 | 2012

Student Seminars

The Department also runs a programme of Research Student Seminars given by PhD students in their 3rd year of study.