Showing papers by "Brno University of Technology published in 2010"

Recurrent neural network based language model

Abstract: A new recurrent neural network based language model (RNN LM) with applications to speech recognition is presented. Results indicate that it is possible to obtain around 50% reduction of perplexity by using mixture of several RNN LMs, compared to a state of the art backoff language model. Speech recognition experiments show around 18% reduction of word error rate on the Wall Street Journal task when comparing models trained on the same amount of data, and around 5% on the much harder NIST RT05 task, even when the backoff model is trained on much more data than the RNN LM. We provide ample empirical evidence to suggest that connectionist language models are superior to standard n-gram techniques, except their high computational (training) complexity. Index Terms: language modeling, recurrent neural networks, speech recognition

An Introduction to Free-space Optical Communications

Abstract: Over the last two decades free-space optical communication (FSO) has become more and more interest- ing as an adjunct or alternative to radio frequency commu- nication. This article gives an overview of the challenges a system designer has to consider while implementing an FSO system. Typical gains and losses along the path from the transmitter through the medium to the receiver are in- troduced in this article. Detailed discussions of these topics can be found in this special issue of the Radioengineering Journal.

Survey on spectrum utilization in Europe: Measurements, analyses and observations

Abstract: This paper presents major results and comparisons of radio spectrum utilization measurements that have been carried out in three different locations in Europe, namely in the suburb of the city of Brno in the Czech Republic and in the suburb and the city of Paris in France during years 2008 and 2009 respectively. The analyzed radio bandwidth extends from 400 MHz to 3 GHz. The measurement method is based on the energy detection principle. Utilization performance and behaviors of major wireless communication systems and primary wireless system within different regions and different environments are investigated here. Our analyses pick out correlations between spectrum utilization within different regions and summarize common observations and physical aspects that will have to be considered in the future radio spectrum management to assure efficient spectrum utilization.

Subspace Gaussian Mixture Models for speech recognition

Abstract: We describe an acoustic modeling approach in which all phonetic states share a common Gaussian Mixture Model structure, and the means and mixture weights vary in a subspace of the total parameter space. We call this a Subspace Gaussian Mixture Model (SGMM). Globally shared parameters define the subspace. This style of acoustic model allows for a much more compact representation and gives better results than a conventional modeling approach, particularly with smaller amounts of training data.

Multilingual acoustic modeling for speech recognition based on subspace Gaussian Mixture Models

Abstract: Although research has previously been done on multilingual speech recognition, it has been found to be very difficult to improve over separately trained systems. The usual approach has been to use some kind of “universal phone set” that covers multiple languages. We report experiments on a different approach to multilingual speech recognition, in which the phone sets are entirely distinct but the model has parameters not tied to specific states that are shared across languages. We use a model called a “Subspace Gaussian Mixture Model” where states' distributions are Gaussian Mixture Models with a common structure, constrained to lie in a subspace of the total parameter space. The parameters that define this subspace can be shared across languages. We obtain substantial WER improvements with this approach, especially with very small amounts of in-language training data.

Parallel genetic algorithm on the CUDA architecture

Abstract: This paper deals with the mapping of the parallel island-based genetic algorithm with unidirectional ring migrations to nVidia CUDA software model. The proposed mapping is tested using Rosenbrock’s, Griewank’s and Michalewicz’s benchmark functions. The obtained results indicate that our approach leads to speedups up to seven thousand times higher compared to one CPU thread while maintaining a reasonable results quality. This clearly shows that GPUs have a potential for acceleration of GAs and allow to solve much complex tasks.

Survey of Period Variations of Superhumps in SU UMa-Type Dwarf Novae. II: The Second Year (2009-2010)

Abstract: As an extension of the project in Kato et al. (2009, arXiv:0905.1757), we collected times of superhump maxima for 61 SU UMa-type dwarf novae mainly observed during the 2009-2010 season. The newly obtained data confirmed the basic findings reported in Kato et al. (2009): the presence of stages A-C, as well as the predominance of positive period derivatives during stage B in systems with superhump periods shorter than 0.07 d. There was a systematic difference in period derivatives for systems with superhump periods longer than 0.075 d between this study and Kato et al. (2009). We suggest that this difference is possibly caused by the relative lack of frequently outbursting SU UMa-type dwarf novae in this period regime in the present study. We recorded a strong beat phenomenon during the 2009 superoutburst of IY UMa. The close correlation between the beat period and superhump period suggests that the changing angular velocity of the apsidal motion of the elliptical disk is responsible for the variation of superhump periods. We also described three new WZ Sge-type objects with established early superhumps and one with likely early superhumps. We also suggest that two systems, VX For and EL UMa, are WZ Sge-type dwarf novae with multiple rebrightenings. The O-C variation in OT J213806.6+261957 suggests that the frequent absence of rebrightenings in very short-Porb objects can be a result of sustained superoutburst plateau at the epoch when usual SU UMa-type dwarf novae return to quiescence preceding a rebrightening. We also present a formulation for a variety of Bayesian extension to traditional period analyses.

Design of a Pressure Sensor Based on Optical Fiber Bragg Grating Lateral Deformation

Abstract: This paper describes steps involved in the design and realization of a new type of pressure sensor based on the optical fiber Bragg grating. A traditional pressure sensor has very limited usage in heavy industrial environments, particularly in explosive or electromagnetically noisy environments. Utilization of optics in these environments eliminates all surrounding influences. An initial motivation for our development was the research, experimental validation, and realization of a complex smart pressure sensor based on the optical principle. The main benefit of this solution consists of increasing sensitivity, resistance to electromagnetic interference, dimensions, and potential increased accuracy.

Total solar eclipse observations of hot prominence shrouds

Abstract: Using observations of the corona taken during the total solar eclipses of 2006 March 29 and 2008 August 1 in broadband white light and in narrow bandpass filters centered at Fe X 637.4 nm, Fe XI 789.2 nm, Fe XIII 1074.7 nm, and Fe XIV 530.3 nm, we show that prominences observed off the solar limb are enshrouded in hot plasmas within twisted magnetic structures. These shrouds, which are commonly referred to as cavities in the literature, are clearly distinct from the overlying arch-like structures that form the base of streamers. The existence of these hot shrouds had been predicted by model studies dating back to the early 1970s, with more recent studies implying their association with twisted magnetic flux ropes. The eclipse observations presented here, which cover a temperature range of 0.9 to 2 ×106 K, are the first to resolve the long-standing ambiguity associated with the temperature and magnetic structure of prominence cavities.

Production of poly(3-hydroxybutyrate- co -3-hydroxyvalerate) by Cupriavidus necator from waste rapeseed oil using propanol as a precursor of 3-hydroxyvalerate

Abstract: Waste rapeseed oil is a useful substrate for polyhydroxyalkanoates (PHA) production employing Cupriavidus necator H16. In fed-batch mode, we obtained biomass and PHA yields of 138 and 105 g l−1, respectively. Yield coefficient and volumetric productivity were 0.83 g PHA per g oil and 1.46 g l−1 h−1, respectively. Propanol at 1% (v/v) enhanced both PHA and biomass formation significantly and, furthermore, resulted in incorporation of 3-hydroxyvalerate units into PHA structure. Thus, propanol can be used as an effective precursor of 3-hydroxyvalarete for production of poly(3-hydroxybutyrate-co-3-hydroxyvalerate) copolymer. During the fed-batch cultivation, propanol concentration was maintained at 1% which resulted in 8% content of 3-hydroxyvalerate in copolymer.

The SYRMEP Beamline of Elettra: Clinical Mammography and Bio-medical Applications

Abstract: At the SYnchrotron Radiation for MEdical Physics (SYRMEP) beamline of Elettra Synchrotron Light Laboratory in Trieste (Italy), an extensive research program in bio‐medical imaging has been developed since 1997. The core program carried out by the SYRMEP collaboration concerns the use of Synchrotron Radiation (SR) for clinical mammography with the aim of improving the diagnostic performance of the conventional technique. The first protocol with patients, started in 2006 has been completed at the end of 2009 and the data analysis is now in progress.Regarding applications different from clinical imaging, synchrotron X‐ray computed microtomography (micro‐CT) is the most used technique, both in absorption and phase contrast. A new software tool, Pore3D, has been developed to perform a quantitative morphological analysis on the reconstructed slices and to access textural information of the sample under study.

Two-Step Sintering of oxide ceramics with various crystal structures

Abstract: The influence of Two-Step Sintering (TSS) process on the final microstructure of oxide ceramic materials with three different crystal structures was studied. Two kinds of alumina (particle size 100 nm resp. 240 nm) as well as tetragonal zirconia (stabilized with 3 mol% Y2O3, particle size 60 nm) and cubic zirconia (8 mol% Y2O3, 140 nm) powders were cold isostatically pressed and pressurelessly sintered with different heating schedules. The microstructures achieved with TSS method were compared with microstructures achieved with conventional Single-Step Sintering (SSS) schedule. The results showed that the efficiency of the TSS of these oxide ceramics was more dependent on their crystal structure than on their particle size and green body microstructure. The method of TSS brought only negligible improvement of the microstructure of tetragonal zirconia and hexagonal alumina ceramics. On the other hand, TSS was successful in the sintering of cubic zirconia ceramics; it led to a decrease in grain size by a factor of 2.

Mutator for transforming memristor into memcapacitor

Abstract: A recently published circuit that operates as a memcapacitor (MC), utilising a memristor (MR) emulator, is analysed with a confirmation that it simulates MC only approximately. A MR to MC mutator is designed by means of off-the-shelf circuits. The SPICE analysis confirms that this circuit transforms exactly the constitutive relation of MR into the constitutive relation of MC.

When simulation meets antichains: on checking language inclusion of nondeterministic finite (tree) automata

Abstract: We describe a new and more efficient algorithm for checking universality and language inclusion on nondeterministic finite word automata (NFA) and tree automata (TA). To the best of our knowledge, the antichain-based approach proposed by De Wulf et al. was the most efficient one so far. Our idea is to exploit a simulation relation on the states of finite automata to accelerate the antichain-based algorithms. Normally, a simulation relation can be obtained fairly efficiently, and it can help the antichain-based approach to prune out a large portion of unnecessary search paths. We evaluate the performance of our new method on NFA/TA obtained from random regular expressions and from the intermediate steps of regular model checking. The results show that our approach significantly outperforms the previous antichain-based approach in most of the experiments.

First-order voltage-mode all-pass filter employing one active element and one grounded capacitor

Abstract: A new circuit topology of first-order voltage-mode all-pass filter providing high-input and low-output impedances is described. The filter consists of only one grounded capacitor and one active element, namely VD-DIBA (Voltage Differencing-Differential Input Buffered Amplifier), with the possibility of electronically tuning the natural frequency. The filter is assembled from commercial integrated circuits, and the frequency responses measured are compared with the theoretical characteristics.

SPICE modelling of memcapacitor

Abstract: A methodology for SPICE modelling of a memcapacitor is described. The model is compounded of the submodel of the state-space system, representing the memory effect of the device, which depends on the concrete physical implementation of the device, and of a capacitor, the capacitance of which is controlled by the system state. The basic fingerprints of the memcapacitor are demonstrated in the example of a charge-controlled memcapacitor.

Coherence-controlled holographic microscope

Abstract: Transmitted-light coherence-controlled holographic microscope (CCHM) based on an off-axis achromatic interferometer allows us to use light sources of arbitrary degree of temporal and spatial coherence. Besides the conventional DHM modes such as quantitative phase contrast imaging and numerical 3D holographic reconstruction it provides high quality (speckle-free) imaging, improved lateral resolution and optical sectioning by coherence gating. Optical setup parameters and their limits for a technical realization are derived and described in detail. To demonstrate the optical sectioning property of the microscope a model sample uncovered and then covered with a diffuser was observed using a low-coherence light source.

An optimized voxel-based morphometric study of gray matter changes in patients with left-sided and right-sided mesial temporal lobe epilepsy and hippocampal sclerosis (MTLE/HS).

Abstract: To determine whether changes in gray matter volume (GMV) differ according to the affected side in mesial temporal lobe epilepsy/hippocampal sclerosis (MTLE/HS) syndrome, and moreover to test the hypothesis of more pronounced structural changes in right-sided MTLE/HS.

Mapping the Distribution of Electron Temperature and Fe Charge States in the Corona with Total Solar Eclipse Observations

Abstract: The inference of electron temperature from the ratio of the intensities of emission lines in the solar corona is valid only when the plasma is collisional. Once collisionless, thermodynamic ionization equilibrium no longer holds, and the inference of an electron temperature and its gradient from such measurements is no longer valid. At the heliocentric distance where the transition from a collision-dominated to a collisionless plasma occurs, the charge states of different elements are established, or frozen-in. These are the charge states which are subsequently measured in interplanetary space. We show in this study how the 2006 March 29 and 2008 August 1 eclipse observations of a number of Fe emission lines yield an empirical value for a distance, which we call Rt , where the emission changes from being collisionally to radiatively dominated. Rt ranges from 1.1 to 2.0 R ☉, depending on the charge state and the underlying coronal density structures. Beyond that distance, the intensity of the emission reflects the distribution of the corresponding Fe ion charge states. These observations thus yield the two-dimensional distribution of electron temperature and charge state measurements in the corona for the first time. The presence of the Fe X 637.4 nm and Fe XI 789.2 nm emission in open magnetic field regions below Rt , such as in coronal holes and the boundaries of streamers, and the absence of Fe XIII 1074.7 nm and Fe XIV 530.3 nm emission there indicate that the sources of the solar wind lie in regions where the electron temperature is less than 1.2 × 106 K. Beyond Rt , the extent of the Fe X [Fe9+] and Fe XI emission [Fe10+], in comparison with Fe XIII [Fe12+] and Fe XIV [Fe13+], matches the dominance of the Fe10+ charge states measured by the Solar Wind Ion Composition Spectrometer, SWICS, on Ulysses, at –43° latitude at 4 AU, in March-April 2006, and Fe9+ and Fe10+ charge states measured by SWICS on the Advanced Composition Explorer, ACE, in the ecliptic plane at 1 AU, at the time of both eclipses. The remarkable correspondence between these two measurements establishes the first direct link between the distribution of charge states in the corona and in interplanetary space.