Transcription of multi-genre media archives using out-of-domain data

We describe our work on developing a speech recognition system for multi-genre media archives. The high diversity of the data makes this a challenging recognition task, which may benefit from systems trained on a combination of in-domain and out-of-domain data. Working with tandem HMMs, we present Multi-level Adaptive Networks (MLAN), a novel technique for incorporating information from out-of-domain posterior features using deep neural networks. We show that it provides a substantial reduction in WER over other systems, with relative WER reductions of 15% over a PLP baseline, 9% over in-domain tandem features and 8% over the best out-of-domain tandem features. © 2012 IEEE.

Sorting nanoparticles by centrifugal fields in clean media

The on-demand availability of nanomaterials with selected size and well-defined chemical/physical properties is of fundamental importance for their widespread application. We report two clean, rapid, and non-destructive approaches for nanoparticle (NP) size selection in centrifugal fields. The first exploits rate zonal separation in a high viscosity gradient. The second exploits selective sedimentation of NPs with different sizes. These methods are here applied to metallic nanoparticles (MNPs) with different compositions and surface chemistry, dispersed either in water or organic solvents. The approach is general and can also be exploited for the separation of NPs of any material. We selectively sort both Au and AgNPs with sizes in the 10-30 nm range, achieving chemical-free MNPs with low polydispersivity. We do not use solutes, thus avoiding contamination, and only require low centrifugal fields, easily achievable in benchtop systems. © 2013 American Chemical Society.

A coupled hydro-mechanical analysis for prediction of hydraulic fracture propagation in saturated porous media using EFG mesh-less method

The details of the Element Free Galerkin (EFG) method are presented with the method being applied to a study on hydraulic fracturing initiation and propagation process in a saturated porous medium using coupled hydro-mechanical numerical modelling. In this EFG method, interpolation (approximation) is based on nodes without using elements and hence an arbitrary discrete fracture path can be modelled.The numerical approach is based upon solving two governing partial differential equations of equilibrium and continuity of pore water simultaneously. Displacement increment and pore water pressure increment are discretized using the same EFG shape functions. An incremental constrained Galerkin weak form is used to create the discrete system of equations and a fully implicit scheme is used for discretization in the time domain. Implementation of essential boundary conditions is based on the penalty method. In order to model discrete fractures, the so-called diffraction method is used.Examples are presented and the results are compared to some closed-form solutions and FEM approximations in order to demonstrate the validity of the developed model and its capabilities. The model is able to take the anisotropy and inhomogeneity of the material into account. The applicability of the model is examined by simulating hydraulic fracture initiation and propagation process from a borehole by injection of fluid. The maximum tensile strength criterion and Mohr-Coulomb shear criterion are used for modelling tensile and shear fracture, respectively. The model successfully simulates the leak-off of fluid from the fracture into the surrounding material. The results indicate the importance of pore fluid pressure in the initiation and propagation pattern of fracture in saturated soils. © 2013 Elsevier Ltd.

Deposition of low stress amorphous zinc tin oxide at ambient temperature using a remote plasma sputtering process suitable for delicate substrates

ZnxSnyOz thin films (<100nm thickness), deposited by remote sputtering from a metal target using a confined argon plasma and oxygen gas jet near the sample, were investigated for their material properties. No visible deformation or curl was observed when deposited on plastic. Materials were confirmed to be amorphous and range between 5 and 10 at.% Sn concentration by x-ray diffraction, x-ray photoemission spectroscopy and energydispersive x-ray spectroscopy. Factors affecting the material composition over time are discussed. Depletion of the Sn as the target ages is suspected. © The Electrochemical Society.

Stabilization of a complex perovskite superstructure under ambient conditions: Influence of cation composition and ordering, and evaluation as an SOFC cathode

Ba1.6Ca2.3Y1.1Fe5O13 is an Fe3+ oxide adopting a complex perovskite superstructure, which is an ordered intergrowth between the Ca2Fe2O5 and YBa2Fe3O8 structures featuring octahedral, square pyramidal, and tetrahedral B sites and three distinct A site environments. The distribution of A site cations was evaluated by combined neutron and X-ray powder diffraction. Consistent with the Fe3+ charge state, the material is an antiferromagnetic insulator with a Néel temperature of 480-485 °C and has a relatively low d.c. conductivity of 2.06 S cm-1 at 700 °C. The observed area specific resistance in symmetrical cell cathodes with the samarium-doped ceria electrolyte is 0.87 Ω cm2 at 700 °C, consistent with the square pyramidal Fe3+ layer favoring oxide ion formation and mobility in the oxygen reduction reaction. Density functional theory calculations reveal factors favoring the observed cation ordering and its influence on the electronic structure, in particular the frontier occupied and unoccupied electronic states. © 2010 American Chemical Society.