MRC Laboratory of Molecular Biology
CUED - Cambridge University Engineering DepartmentNot a member yet
45551 research outputs found
Sort by
Conversionless efficient and broadband laser light diffusers for high brightness illumination applications.
No full textLaser diodes are efficient light sources. However, state-of-the-art laser diode-based lighting systems rely on light-converting inorganic phosphor materials, which strongly limit the efficiency and lifetime, as well as achievable light output due to energy losses, saturation, thermal degradation, and low irradiance levels. Here, we demonstrate a macroscopically expanded, three-dimensional diffuser composed of interconnected hollow hexagonal boron nitride microtubes with nanoscopic wall-thickness, acting as an artificial solid fog, capable of withstanding ~10 times the irradiance level of remote phosphors. In contrast to phosphors, no light conversion is required as the diffuser relies solely on strong broadband (full visible range) lossless multiple light scattering events, enabled by a highly porous (>99.99%) non-absorbing nanoarchitecture, resulting in efficiencies of ~98%. This can unleash the potential of lasers for high-brightness lighting applications, such as automotive headlights, projection technology or lighting for large spaces
Universal adversarial attacks on spoken language assessment systems
No full textThere is an increasing demand for automated spoken language assessment (SLA) systems, partly driven by the performance improvements that have come from deep learning based approaches. One aspect of deep learning systems is that they do not require expert derived features, operating directly on the original signal such as a speech recognition (ASR) transcript. This, however, increases their potential susceptibility to adversarial attacks as a form of candidate malpractice. In this paper the sensitivity of SLA systems to a universal black-box attack on the ASR text output is explored. The aim is to obtain a single, universal phrase to maximally increase any candidate's score. Four approaches to detect such adversarial attacks are also described. All the systems, and associated detection approaches, are evaluated on a free (spontaneous) speaking section from a Business English test. It is shown that on deep learning based SLA systems the average candidate score can be increased by almost one grade level using a single six word phrase appended to the end of the response hypothesis. Although these large gains can be obtained, they can be easily detected based on detection shifts from the scores of a “traditional” Gaussian Process based grader
Molecular level simulations of combustion processes using the DSMC method
No full text© 2021 Informa UK Limited, trading as Taylor & Francis Group. A technique called the Direct Simulation Monte Carlo (DSMC) method is used for simulating laminar one-dimensional hydrogen-air flames with an aim to establish the feasibility of molecular-level simulations for combustion using the Quantum-Kinetic (QK) reaction model. In DSMC, simulation particles are employed which represent many molecules with similar properties. The DSMC method effectively solves the Boltzmann equation by decoupling the motion of molecules into two phases: a Move phase and a Collision phase. Chemistry is treated using the Quantum-Kinetic model in which the total energy exchange resulting from molecular collision determines the outcome of the chemical reactions. A major advantage of DSMC is that it avoids using continuum Arrhenius reaction rates or simplified gradient laws for the diffusivities of mass, momentum and heat. Instead, any such laws and their parameters emerge naturally from the DSMC results. Simulations of one-dimensional hydrogen-air flames using the DSMC method with a detailed chemical scheme show promising results for molecular-level simulations of combustion. An adjustment to the activation energy of a key reaction is made in order to achieve the correct flame speed. The species and temperature profiles from DSMC show reasonable agreement with those obtained from Direct Numerical Simulation (DNS) using a conventional continuum approach. Likewise, the molecular diffusivity of hydrogen and oxygen also show reasonable agreement with the DNS results, although differences in the results still exist, especially for oxygen diffusivity. These are expected to improve with further development in the DSMC method for combustion
Ion conductivity through TEMPO-mediated oxidated and periodate oxidated cellulose membranes
No full textCellulose in different forms is increasingly used due to sustainability aspects. Even though cellulose itself is an isolating material, it might affect ion transport in electronic applications. This effect is important to understand for instance in the design of cellulose-based supercapacitors. To test the ion conductivity through membranes made from cellulose nanofibril (CNF) materials, different electrolytes chosen with respect to the Hofmeister series were studied. The CNF samples were oxidised to three different surface charge levels via 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO), and a second batch was further cross-linked by periodate oxidation to increase wet strength and stability. The outcome showed that the CNF pre-treatment and choice of electrolyte are both crucial to the ion conductivity through the membranes. Significant specific ion effects were observed for the TEMPO-oxidised CNF. Periodate oxidated CNF showed low ion conductivity for all electrolytes tested due to an inhibited swelling caused by the crosslinking reaction
HoloBlade: An open platform for holography
No full textWe present HoloBlade, a collection of open and accessible technologies to advance the field of holography and serve the wider research community. The initial implementation is presented and functionality demonstrated
Near Magnetic Field Emission Analysis for IGBT and SiC Power Modules
No full textPower modules with high-speed power devices such as IGBTs and SiC MOSFETs have become crucial components of medium to high power applications. However, their faster switching speed and compact design raise near magnetic field emission issues. The near magnetic field from the high-speed power modules could contaminate the peripheral circuits and filters. In this paper, the near magnetic field emission is analyzed for three power modules. The root cause and sources of the near magnetic emission are identified. Based on the analysis, the near magnetic field is predicted and verified with finite element analysis (FEA) simulation and measurement
Internet of Radars: Sensing versus Sending with Joint Radar-Communications
No full textThe Internet of Things (IoT) is made up of interconnected devices for exchanging information through sensors and actuators. One of the main physical sensors to understand the environment beyond the visible world is a radar. Basically, radars have always been a military tool to investigate the environment. However, with the developing technology, radars have become more compact and affordable to use in a building, in a car, in a drone, or even in a wristwatch. In the near future, radar-equipped IoT platforms will start to appear increasingly. For each IoT platform, dual use of spectrum with dual aperture is required for sensing and communicating when using conventional approaches. Emissions from the radar and communication circuitries are the main causes of the increase in energy consumption for any radar sensing IoT device. Furthermore, an increasing number of radars cause congested spectrum, and RF convergence between radar and communication systems becomes more likely to present itself. In recent years, numerous research works have proposed using the single waveform for perceiving the environment and sending information. They are often called "joint radar-communication' (JRC) systems. As a result of the latest advancements in JRC system designs, radar sensing IoT platforms now can be transformed into an "Internet of Radars"(IoR). This article is an attempt to introduce a prospective research direction in order to develop the architectures necessary to make the IoR concept possible. In this article, we present a short survey of JRC technologies likely to be used on radar-sensing-capable IoT devices. Then possible application areas, challenges to enable JRC, and future research perspectives are proposed
Open innovation environments as knowledge sharing enablers: the case of strategic technology and innovative management consortium
No full textPurpose: This study aims to understand how open innovation (OI) environments can help organizations in implementing knowledge sharing (KS) practices defusing KS barriers. Design/methodology/approach: An in-depth case study analysis on the strategic technology and innovation management (STIM) consortium at the Institute of Manufacturing of the University of Cambridge was performed during the 2019 and 2020 STIM program editions. To analyze data, this paper used the interpretive structural model on a sample of 20 managers participating in the STIM consortium, and this paper carried out an exploratory in-depth case study analysis to validate the results. Findings: The findings shed light on the role of OI environments in defusing KS barriers in the process of inter-organizational KS. Originality/value: Notwithstanding the importance of KS practices among organizations, only a few studies have recognized and investigated the role played by OI arrangements in enhancing KS practices
Predicting semantic map representations from images using pyramid occupancy networks
No full textAutonomous vehicles commonly rely on highly detailed birds-eye-view maps of their environment, which capture both static elements of the scene such as road layout as well as dynamic elements such as other cars and pedestrians. Generating these map representations on the fly is a complex multi-stage process which incorporates many important vision-based elements, including ground plane estimation, road segmentation and 3D object detection. In this work we present a simple, unified approach for estimating these map representations directly from monocular images using a single end-to-end deep learning architecture. For the maps themselves we adopt a semantic Bayesian occupancy grid framework, allowing us to trivially accumulate information over multiple cameras and timesteps. We demonstrate the effectiveness of our approach by evaluating against several challenging baselines on the NuScenes and Argoverse datasets, and show that we are able to achieve a relative improvement of 9.1% and 22.3% respectively compared to the best-performing existing method
Novel and simple patterning process of quantum dots via transfer printing for active matrix qd-led
No full text© 2020 SID. The next generation of a self-emitting display requires precise and stable patterning techniques to shape Red, Green, and Blue pixels using quantum dots. In this study, we propose the novel and simple transfer printing process for the active matrix QD-LEDs