Multistream Articulatory Feature-Based Models for Visual Speech Recognition

We study the problem of automatic visual speech recognition (VSR) using dynamic Bayesian network (DBN)-based models consisting of multiple sequences of hidden states, each corresponding to an articulatory feature (AF) such as lip opening (LO) or lip rounding (LR). A bank of discriminative articulatory feature classifiers provides input to the DBN, in the form of either virtual evidence (VE) (scaled likelihoods) or raw classifier margin outputs. We present experiments on two tasks, a medium-vocabulary word-ranking task and a small-vocabulary phrase recognition task. We show that articulatory feature-based models outperform baseline models, and we study several aspects of the models, such as the effects of allowing articulatory asynchrony, of using dictionary-based versus whole-word models, and of incorporating classifier outputs via virtual evidence versus alternative observation models.     

Right-first-timendashthe role of automatic dispensing in the dyehouse

Adding dyes and chemicals accurately to the dyebath is critical in making dyeing right-first-time. Trevor Larkins looks at how automatic dispensing can help.     

ESMRMB 2017, 34th Annual Scientific Meeting,Barcelona, ES,October 19–October 21: Abstracts,Thursday

Objective

Current magnetic resonance imaging (MRI) axon diameter measurements rely on the pulsed gradient spin-echo sequence, which is unable to provide diffusion times short enough to measure small axon diameters. This study combines the AxCaliber axon diameter fitting method with data generated from Monte Carlo simulations of oscillating gradient spin-echo sequences (OGSE) to infer micron-sized axon diameters, in order to determine the feasibility of using MRI to infer smaller axon diameters in brain tissue.

Materials and methods

Monte Carlo computer simulation data were synthesized from tissue geometries of cylinders of different diameters using a range of gradient frequencies in the cosine OGSE sequence . Data were fitted to the AxCaliber method modified to allow the new pulse sequence. Intra- and extra-axonal water were studied separately and together.

Results

The simulations revealed the extra-axonal model to be problematic. Rather than change the model, we found that restricting the range of gradient frequencies such that the measured apparent diffusion coefficient was constant over that range resulted in more accurate fitted diameters. Thus a careful selection of frequency ranges is needed for the AxCaliber method to correctly model extra-axonal water, or adaptations to the method are needed. This restriction helped reduce the necessary gradient strengths for measurements that could be performed with parameters feasible for a Bruker BG6 gradient set. For these experiments, the simulations inferred diameters as small as 0.5 μm on square-packed and randomly packed cylinders. The accuracy of the inferred diameters was found to be dependent on the signal-to-noise ratio (SNR), with smaller diameters more affected by noise, although all diameter distributions were distinguishable from one another for all SNRs tested.

Conclusion

The results of this study indicate the feasibility of using MRI with OGSE on preclinical scanners to infer small axon diameters.

     

Response of reinforced concrete beams to hydrostatic pressure acting within primary cracks

This paper investigates the response of reinforced concrete beams to hydrostatic pressure acting within primary cracks. Notched beams were initially pre-cracked before pressurised water was introduced into the primary crack. The deflection and strain increase at tensile reinforcement level (due to the pressurised water load) was measured. Tests were carried out using applied hydrostatic pressures of 0.2 and 0.325 MPa. Results show that both deflection and strain at tensile reinforcement level increases immediately after the introduction of hydrostatic pressure into open primary cracks. If the crack is held open and hydrostatic pressure is allowed to build up within the depth of the crack, additional deformation occurs. A finite element (FE) model was constructed to investigate the effects of greater hydrostatic pressures acting within the primary crack. The FE model was first validated against the test data, before being used to assess the structural response of the reinforced concrete section to applied hydrostatic pressures of up to 1 MPa. It was found that section deformations increased as hydrostatic pressure was increased. At applied hydrostatic pressures of 0.8 MPa and above, the increases in strain at tensile reinforcement level were shown to be significant. DNV-0S-C205, which is the Det Norske Veritas (DNV) standard for offshore concrete structures states that the ‘effects of water pressure within cracks may be neglected for structural elements exposed to less than 100 m (0.981 MPa) of water head.’ The current research suggests that the effects of water pressure within cracks for structural elements exposed to slightly less than 100 m water head may also be significant (based on a 10 % threshold criteria). However, it is accepted that a more comprehensive parametric study would be required to determine whether or not the DNV should be redressed.     

Mixed mode fracture behaviour of steel fibre reinforced concrete

In this paper, results are reported for a series of discrete end hooked and straight fibre pullout tests subjected to mixed mode action with the results compared to that of discrete fibres pulled out in Mode I (tensile) and Mode II (shear) fracture. As has been previously observed from Modes I and II fracture tests, the snubbing effect dominates the behaviour of fibres at large fibre bending angles. At large fibre bending angles, considerable slip and crack separation occurred prior to the fibres being engaged in taking load and fibres that are inclined close to the cracked surface are ineffective in carrying load. The results of the test were compared with the fibre engagement and bond stress models in the Unified Variable Engagement Model (UVEM). A good correlation is observed for the UVEM model with the test data and provides further confirmation of the validity of the UVEM model to predict the mix mode fracture of steel fibre reinforced concrete.     

Low emittance coatings and the thermal performance of vacuum glazing

The thermal performances of vacuum glazings employing coatings with emittance between 0.02 and 0.16 were simulated using a three-dimensional finite volume model. Physical samples of vacuum glazings with hard and soft coatings with emittance of 0.04, 0.12 and 0.16 were fabricated and their thermal performance characterised experimentally using a guarded hot box calorimeter. Good agreement was found between experimental and theoretical thermal performances for both a vacuum glazing with a soft coating (emittance 0.04) and those with hard coatings (emittance 0.12 and 0.16). Simulations showed that for a low value of emittance (e.g. 0.02), the use of two low-emittance coatings gives limited improvement in thermal performance of the glazing system. The use of a single high performance low-emittance coating in a vacuum glazing has been shown to provide excellent performance.     

Experimental validation of a numerical model for heat transfer in vacuum glazing

Flat vacuum glazings consisting of a narrow evacuated space between two glass panes separated by an array of small support pillars have been fabricated. A guarded hot box calorimeter was designed and constructed to measure their heat transfer coefficients. Experimental measurements of temperatures and rates of heat transfer were found to be in very good agreement with those predicted using a developed finite element model. A method for determining the heat transfer coefficient of the evacuated gap has been established and comparisons are made between the measured and predicted glass surface temperature profiles of the exposed glass area and the heat transfer coefficients of the total glazing system in order to validated the model.     

Simple modelling of the effects of exploitation on hot springs, Geyser Valley, Wairakei, New Zealand

Many hot springs in Geyser Valley declined and ceased flowing during well testing (1950–1958) or the early stages of development (1958–1964) of the Wairakei field, North Island, New Zealand. A simple model that considers mixing, geothermal reservoir pressure, and Darcy flow is applied to analyse the changes in flow rate and chemistry of spring SP18. The model suggests that prior to the testing the observed flow rate of 3 l/s was comprised of about 2.1 l/s of water from the deep reservoir and about 0.9 l/s from shallow groundwater. As the pressure in the reservoir decreased as a result of exploitation, the reservoir component declined but the groundwater component remained near constant until the spring had nearly ceased flowing. The model reasonably predicts the flow cessation date for four other springs (Group A) but poorly predicts that for seven other springs (Group B). These two groups of springs do not appear to be clustered at the ground surface. The pressure–elevation relationship suggests that the Group A springs feed from within the shallower Waiora Formation and Group B springs from within the deeper Wairakei Ignimbrite Formation. The observation that declines of chloride concentrations in Group A springs occurred at least three years before those in springs of Group B is consistent with the finding that Group A springs have a shallower feeder depth.     

Improving the Reliability of MLC NAND Flash Memories Through Adaptive Data Refresh and Error Control Coding

NAND Flash memory has become the most widely used non-volatile memory technology. We focus on multi-level cell (MLC) NAND Flash memories because they have high storage density. Unfortunately MLC NAND Flash memory also has reliability problems due to narrower threshold voltage gap between logical states. Errors in these memories can be classified into data retention (DR) errors and program interference (PI) errors. DR errors are dominant if the data storage time is longer than 1 day and these errors can be reduced by refreshing the data. PI errors are dominant if the data storage time is less than 1 day and these errors can be handled by error control coding (ECC). In this paper we propose a combination of data refresh policies and low cost ECC schemes that are cognizant of application characteristics to address the errors in MLC NAND Flash memories. First, we use Gray code based encoding to reduce the error rates in the four subpages (MSB-even, LSB-even, MSB-odd, LSB-odd) of a 2-bit MLC NAND Flash memory. Next, we apply data refresh techniques where the refresh interval is a function of the program/erase (P/E) frequency of the application. We show that an appropriate choice of refresh interval and BCH based ECC scheme can minimize memory energy while satisfying the reliability constraint.