Showing papers by "Brno University of Technology published in 2016"

Plasma-liquid interactions: A review and roadmap

Abstract: Plasma–liquid interactions represent a growing interdisciplinary area of research involving plasma science, fluid dynamics, heat and mass transfer, photolysis, multiphase chemistry and aerosol science. This review provides an assessment of the state-of-the-art of this multidisciplinary area and identifies the key research challenges. The developments in diagnostics, modeling and further extensions of cross section and reaction rate databases that are necessary to address these challenges are discussed. The review focusses on non-equilibrium plasmas.

BoxCars: 3D Boxes as CNN Input for Improved Fine-Grained Vehicle Recognition

Abstract: We are dealing with the problem of fine-grained vehicle make&model recognition and verification. Our contribution is showing that extracting additional data from the video stream – besides the vehicle image itself – and feeding it into the deep convolutional neural network boosts the recognition performance considerably. This additional information includes: 3D vehicle bounding box used for "unpacking" the vehicle image, its rasterized low-resolution shape, and information about the 3D vehicle orientation. Experiments show that adding such information decreases classification error by 26% (the accuracy is improved from 0.772 to 0.832) and boosts verification average precision by 208% (0.378 to 0.785) compared to baseline pure CNN without any input modifications. Also, the pure baseline CNN outperforms the recent state of the art solution by 0.081. We provide an annotated set "BoxCars" of surveillance vehicle images augmented by various automatically extracted auxiliary information. Our approach and the dataset can considerably improve the performance of traffic surveillance systems.

Design of power-efficient approximate multipliers for approximate artificial neural networks

Abstract: Artificial neural networks (NN) have shown a significant promise in difficult tasks like image classification or speech recognition. Even well-optimized hardware implementations of digital NNs show significant power consumption. It is mainly due to non-uniform pipeline structures and inherent redundancy of numerous arithmetic operations that have to be performed to produce each single output vector. This paper provides a methodology for the design of well-optimized power-efficient NNs with a uniform structure suitable for hardware implementation. An error resilience analysis was performed in order to determine key constraints for the design of approximate multipliers that are employed in the resulting structure of NN. By means of a search based approximation method, approximate multipliers showing desired tradeoffs between the accuracy and implementation cost were created. Resulting approximate NNs, containing the approximate multipliers, were evaluated using standard benchmarks (MNIST dataset) and a real-world classification problem of Street-View House Numbers. Significant improvement in power efficiency was obtained in both cases with respect to regular NNs. In some cases, 91% power reduction of multiplication led to classification accuracy degradation of less than 2.80%. Moreover, the paper showed the capability of the back propagation learning algorithm to adapt with NNs containing the approximate multipliers.

Coordination Chemistry Dictates the Structural Defects in Lead Halide Perovskites

Abstract: We show the influence of species present in precursor solution during formation of lead halide perovskite materials on the structural defects of the films. The coordination of lead by competing solvent molecules and iodide ions dictate the type of complexes present in the films. Depending on the processing conditions all PbIS5+, PbI2S4, PbI3S3−, PbI4S22−, PbI5S23−, PbI64−and 1D (Pb2I4)n chains are observed by absorption measurements. Different parameters are studied such as polarity of the solvent, concentration of iodide ions, concentration of solvent molecules and temperature. It is concluded that strongly coordinating solvents will preferentially form species with a low number of iodide ions and less coordinative solvents generate high concentration of PbI6−. We furthermore propose that all these plumbate ions may act as structural defects determining electronic properties of the photovoltaic films.

Vehicle Re-identification for Automatic Video Traffic Surveillance

Abstract: This paper proposes an approach to the vehicle reidentification problem in a multiple camera system. We focused on the re-identification itself assuming that the vehicle detection problem is already solved including extraction of a full-fledged 3D bounding box. The re-identification problem is solved by using color histograms and histograms of oriented gradients by a linear regressor. The features are used in separate models in order to get the best results in the shortest CPU computation time. The proposed method works with a high accuracy (60% true positives retrieved with 10% false positive rate on a challenging subset of the test data) in 85 milliseconds of the CPU (Core i7) computation time per one vehicle re-identification assuming the fullHD resolution video input. The applications of this work include finding important parameters such as travel time, traffic flow, or traffic information in a distributed traffic surveillance and monitoring system.

PredictSNP2: A Unified Platform for Accurately Evaluating SNP Effects by Exploiting the Different Characteristics of Variants in Distinct Genomic Regions.

Abstract: An important message taken from human genome sequencing projects is that the human population exhibits approximately 99.9% genetic similarity. Variations in the remaining parts of the genome determine our identity, trace our history and reveal our heritage. The precise delineation of phenotypically causal variants plays a key role in providing accurate personalized diagnosis, prognosis, and treatment of inherited diseases. Several computational methods for achieving such delineation have been reported recently. However, their ability to pinpoint potentially deleterious variants is limited by the fact that their mechanisms of prediction do not account for the existence of different categories of variants. Consequently, their output is biased towards the variant categories that are most strongly represented in the variant databases. Moreover, most such methods provide numeric scores but not binary predictions of the deleteriousness of variants or confidence scores that would be more easily understood by users. We have constructed three datasets covering different types of disease-related variants, which were divided across five categories: (i) regulatory, (ii) splicing, (iii) missense, (iv) synonymous, and (v) nonsense variants. These datasets were used to develop category-optimal decision thresholds and to evaluate six tools for variant prioritization: CADD, DANN, FATHMM, FitCons, FunSeq2 and GWAVA. This evaluation revealed some important advantages of the category-based approach. The results obtained with the five best-performing tools were then combined into a consensus score. Additional comparative analyses showed that in the case of missense variations, protein-based predictors perform better than DNA sequence-based predictors. A user-friendly web interface was developed that provides easy access to the five tools’ predictions, and their consensus scores, in a user-understandable format tailored to the specific features of different categories of variations. To enable comprehensive evaluation of variants, the predictions are complemented with annotations from eight databases. The web server is freely available to the community at http://loschmidt.chemi.muni.cz/predictsnp2.

PMSM Model Predictive Control With Field-Weakening Implementation

Abstract: Permanent-magnet synchronous machine (PMSM) drives have become popular for motion control applications due to their performance and high torque-to-weight ratio. The complex task of PMSM control in high-performance applications is currently usually resolved with classical vector control. Modern control techniques such as model predictive control (MPC) can provide significant benefits over field-oriented control, especially with straightforward controller tuning and constraints handling. Unfortunately, these new algorithms usually suffer from problems with their computational complexity. In this paper, we present a PMSM control method based on an explicit MPC with a novel linearization and constraints handling method, allowing natural field weakening. The algorithm was designed with respect to computationally feasible implementation in the control hardware. The proposed control algorithm has been proved and successfully verified in both simulation and implementation on a real motor.

Analysis of DNN approaches to speaker identification

Abstract: This work studies the usage of the Deep Neural Network (DNN) Bottleneck (BN) features together with the traditional MFCC features in the task of i-vector-based speaker recognition We decouple the sufficient statistics extraction by using separate GMM models for frame alignment, and for statistics normalization and we analyze the usage of BN and MFCC features (and their concatenation) in the two stages We also show the effect of using full-covariance GMM models, and, as a contrast, we compare the result to the recent DNN-alignment approach On the NIST SRE2010, telephone condition, we show 60% relative gain over the traditional MFCC baseline for EER (and similar for the NIST DCF metrics), resulting in 094% EER

Simple Chaotic Flow with Circle and Square Equilibrium

Abstract: Simple systems of third-order autonomous nonlinear differential equations can exhibit chaotic behavior. In this paper, we present a new class of chaotic flow with a square-shaped equilibrium. This unique property has apparently not yet been described. Such a system belongs to a newly introduced category of chaotic systems with hidden attractors that are interesting and important in engineering applications. The mathematical model is accompanied by an electrical circuit implementation, demonstrating structural stability of the strange attractor. The circuit is simulated with PSpice, constructed, and analyzed (measured).

Magneto-Optical Signature of Massless Kane Electrons in Cd3As2

Abstract: We report on optical reflectivity experiments performed on Cd3As2 over a broad range of photon energies and magnetic fields. The observed response clearly indicates the presence of 3D massless charge carriers. The specific cyclotron resonance absorption in the quantum limit implies that we are probing massless Kane electrons rather than symmetry-protected 3D Dirac particles. The latter may appear at a smaller energy scale and are not directly observed in our infrared experiments.

Aerosol-Assisted CVD-Grown PdO Nanoparticle-Decorated Tungsten Oxide Nanoneedles Extremely Sensitive and Selective to Hydrogen.

Abstract: We report for the first time the successful synthesis of palladium (Pd) nanoparticle (NP)-decorated tungsten trioxide (WO3) nanoneedles (NNs) via a two-step aerosol-assisted chemical vapor deposition approach. Morphological, structural, and elemental composition analysis revealed that a Pd(acac)2 precursor was very suitable to decorate WO3 NNs with uniform and well-dispersed PdO NPs. Gas-sensing results revealed that decoration with PdO NPs led to an ultrasensitive and selective hydrogen (H2) gas sensor (sensor response peaks at 1670 at 500 ppm of H2) with low operating temperature (150 °C). The response of decorated NNs is 755 times higher than that of bare WO3 NNs. Additionally, at a temperature near that of the ambient temperature (50 °C), the response of this sensor toward the same concentration of H2 was 23, which is higher than that of some promising sensors reported in the literature. Finally, humidity measurements showed that PdO/WO3 sensors displayed low-cross-sensitivity toward water vapor, compared to bare WO3 sensors. The addition of PdO NPs helps to minimize the effect of ambient humidity on the sensor response.

Collar Line Segments for fast odometry estimation from Velodyne point clouds

Abstract: We present a novel way of odometry estimation from Velodyne LiDAR point cloud scans. The aim of our work is to overcome the most painful issues of Velodyne data - the sparsity and the quantity of data points - in an efficient way, enabling more precise registration. Alignment of the point clouds which yields the final odometry is based on random sampling of the clouds using Collar Line Segments (CLS). The closest line segment pairs are identified in two sets of line segments obtained from two consequent Velodyne scans. From each pair of correspondences, a transformation aligning the matched line segments into a 3D plane is estimated. By this, significant planes (ground, walls, …) are preserved among aligned point clouds. Evaluation using the KITTI dataset shows that our method outperforms publicly available and commonly used state-of-the-art method GICP for point cloud registration in both accuracy and speed, especially in cases where the scene lacks significant landmarks or in typical urban elements. For such environments, the registration error of our method is reduced by 75% compared to the original GICP error.

Exploring synergy between communications, caching, and computing in 5G-grade deployments

Abstract: Decisive progress in 5G mobile technology, fueled by a rapid proliferation of computation- hungry and delay-sensitive services, puts economic pressure on the research community to rethink the fundamentals of underlying networking architectures. Along these lines, the first half of this article offers a first-hand tutorial on the most recent advances in content-centric networking, emerging user applications, as well as enabling system architectures. We establish that while significant progress has been made along the individual vectors of communications, caching, and computing, together with some promising steps in proposing hybrid functionalities, the ultimate synergy behind a fully integrated solution is not nearly well understood. Against this background, the second half of this work carefully brings into perspective additional important factors, such as user mobility patterns, aggressive application requirements, and associated operator deployment capabilities, to conduct comprehensive system-level analysis. Furthermore, supported by a full-fledged practical trial on a live cellular network, our systematic findings reveal the most dominant factors in converged 5G-grade communications, caching, and computing layouts, as well as indicate the natural optimization points for system operators to leverage the maximum available benefits.

Sequence summarizing neural network for speaker adaptation

Abstract: In this paper, we propose a DNN adaptation technique, where the i-vector extractor is replaced by a Sequence Summarizing Neural Network (SSNN). Similarly to i-vector extractor, the SSNN produces a "summary vector", representing an acoustic summary of an utterance. Such vector is then appended to the input of main network, while both networks are trained together optimizing single loss function. Both the i-vector and SSNN speaker adaptation methods are compared on AMI meeting data. The results show comparable performance of both techniques on FBANK system with frame-classification training. Moreover, appending both the i-vector and "summary vector" to the FBANK features leads to additional improvement comparable to the performance of FMLLR adapted DNN system.

Accumulation of Poly(3-hydroxybutyrate) Helps Bacterial Cells to Survive Freezing

Abstract: Accumulation of polyhydroxybutyrate (PHB) seems to be a common metabolic strategy adopted by many bacteria to cope with cold environments. This work aimed at evaluating and understanding the cryoprotective effect of PHB. At first a monomer of PHB, 3-hydroxybutyrate, was identified as a potent cryoprotectant capable of protecting model enzyme (lipase), yeast (Saccharomyces cerevisiae) and bacterial cells (Cupriavidus necator) against the adverse effects of freezing-thawing cycles. Further, the viability of the frozen–thawed PHB accumulating strain of C. necator was compared to that of the PHB non-accumulating mutant. The presence of PHB granules in cells was revealed to be a significant advantage during freezing. This might be attributed to the higher intracellular level of 3-hydroxybutyrate in PHB accumulating cells (due to the action of parallel PHB synthesis and degradation, the so-called PHB cycle), but the cryoprotective effect of PHB granules seems to be more complex. Since intracellular PHB granules retain highly flexible properties even at extremely low temperatures (observed by cryo-SEM), it can be expected that PHB granules protect cells against injury from extracellular ice. Finally, thermal analysis indicates that PHB-containing cells exhibit a higher rate of transmembrane water transport, which protects cells against the formation of intracellular ice which usually has fatal consequences.

Evaluation of 3-hydroxybutyrate as an enzyme-protective agent against heating and oxidative damage and its potential role in stress response of poly(3-hydroxybutyrate) accumulating cells

Abstract: Poly(3-hydroxybutyrate) (PHB) is a common carbon- and energy-storage compound simultaneously produced and degraded into its monomer 3-hydroxybutyrate (3HB) by numerous bacteria and Archae in a metabolic pathway called the PHB cycle. We investigated 3HB as a chemical chaperone capable of protecting model enzymes, namely lipase and lysozyme, from adverse effects of high temperature and oxidation. Heat-mediated denaturation of lipase in the presence or absence of 3HB was monitored by dynamic light scattering (DLS) revealing a significant protective effect of 3HB which increased as its concentration rose. Furthermore, when compared at the same molar concentration, 3HB showed a greater protective effect than the well-known chemical chaperones trehalose and hydroxyectoine. The higher protective effect of 3HB was also confirmed when employing differential scanning calorimetry (DSC) and lysozyme as a model enzyme. Furthermore, 3HB was capable of protecting lipase not only against thermal-mediated denaturation but also against oxidative damage by Cu(2+) and H2O2; its protection was higher than that of trehalose and comparable to that of hydroxyectoine. Taking into account that the PHB-producing strain Cupriavidus necator H16 reveals a 16.5-fold higher intracellular concentration than the PHB non-producing mutant C. necator PHB(-4), it might be expected that the functional PHB cycle might be responsible for maintaining a higher intracellular level of 3HB which, aside from other positive aspects of functional PHB metabolism, enhances stress resistance of bacterial strains capable of simultaneous PHB synthesis and mobilization. In addition, 3HB can be used in various applications and formulations as an efficient enzyme-stabilizing and enzyme-protecting additive.

Feasibility characterization of cryptographic primitives for constrained (wearable) IoT devices

Abstract: The Internet of Things (IoT) employs smart devices as its building blocks for developing a ubiquitous communication framework. It thus supports a wide variety of application domains, including public safety, healthcare, education, and public transportation. While offering a novel communication paradigm, IoT finds its requirements closely connected to the security issues. The role of security following the fact that a new type of devices known as wearables constitute an emerging area. This paper delivers an applicability study of the state-of-the-art cryptographic primitives for wearable IoT devices, including the pairing-based cryptography. Pairing-based schemes are well-recognized as fundamental enablers for many advanced cryptographic applications, such as privacy protection and identity-based encryption. To deliver a comprehensive view on the computational power of modern wearable devices (smart phones, watches, and embedded devices), we perform an evaluation of a variety of them utilizing bilinear pairing for real-time communication. In order to deliver a complete picture, the obtained bilinear pairing results are complemented with performance figures for classical cryptography (such as block ciphers, digital signatures, and hash functions). Our findings show that wearable devices of today have the needed potential to efficiently operate with cryptographic primitives in real time. Therefore, we believe that the data provided during this research would shed light on what devices are more suitable for certain cryptographic operations.

HotSpot Wizard 2.0: automated design of site-specific mutations and smart libraries in protein engineering

Abstract: HotSpot Wizard 2.0 is a web server for automated identification of hot spots and design of smart libraries for engineering proteins' stability, catalytic activity, substrate specificity and enantioselectivity. The server integrates sequence, structural and evolutionary information obtained from 3 databases and 20 computational tools. Users are guided through the processes of selecting hot spots using four different protein engineering strategies and optimizing the resulting library's size by narrowing down a set of substitutions at individual randomized positions. The only required input is a query protein structure. The results of the calculations are mapped onto the protein's structure and visualized with a JSmol applet. HotSpot Wizard lists annotated residues suitable for mutagenesis and can automatically design appropriate codons for each implemented strategy. Overall, HotSpot Wizard provides comprehensive annotations of protein structures and assists protein engineers with the rational design of site-specific mutations and focused libraries. It is freely available at http://loschmidt.chemi.muni.cz/hotspotwizard.