Showing papers by "Mitsubishi Electric published in 2017"

CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos

Abstract: Temporal action localization is an important yet challenging problem. Given a long, untrimmed video consisting of multiple action instances and complex background contents, we need not only to recognize their action categories, but also to localize the start time and end time of each instance. Many state-of-the-art systems use segment-level classifiers to select and rank proposal segments of pre-determined boundaries. However, a desirable model should move beyond segment-level and make dense predictions at a fine granularity in time to determine precise temporal boundaries. To this end, we design a novel Convolutional-De-Convolutional (CDC) network that places CDC filters on top of 3D ConvNets, which have been shown to be effective for abstracting action semantics but reduce the temporal length of the input data. The proposed CDC filter performs the required temporal upsampling and spatial downsampling operations simultaneously to predict actions at the frame-level granularity. It is unique in jointly modeling action semantics in space-time and fine-grained temporal dynamics. We train the CDC network in an end-to-end manner efficiently. Our model not only achieves superior performance in detecting actions in every frame, but also significantly boosts the precision of localizing temporal boundaries. Finally, the CDC network demonstrates a very high efficiency with the ability to process 500 frames per second on a single GPU server. Source code and trained models are available online at https://bitbucket.org/columbiadvmm/cdc.

CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos

Abstract: Temporal action localization is an important yet challenging problem. Given a long, untrimmed video consisting of multiple action instances and complex background contents, we need not only to recognize their action categories, but also to localize the start time and end time of each instance. Many state-of-the-art systems use segment-level classifiers to select and rank proposal segments of pre-determined boundaries. However, a desirable model should move beyond segment-level and make dense predictions at a fine granularity in time to determine precise temporal boundaries. To this end, we design a novel Convolutional-De-Convolutional (CDC) network that places CDC filters on top of 3D ConvNets, which have been shown to be effective for abstracting action semantics but reduce the temporal length of the input data. The proposed CDC filter performs the required temporal upsampling and spatial downsampling operations simultaneously to predict actions at the frame-level granularity. It is unique in jointly modeling action semantics in space-time and fine-grained temporal dynamics. We train the CDC network in an end-to-end manner efficiently. Our model not only achieves superior performance in detecting actions in every frame, but also significantly boosts the precision of localizing temporal boundaries. Finally, the CDC network demonstrates a very high efficiency with the ability to process 500 frames per second on a single GPU server. We will update the camera-ready version and publish the source codes online soon.

FoldingNet: Point Cloud Auto-encoder via Deep Grid Deformation

Abstract: Recent deep networks that directly handle points in a point set, e.g., PointNet, have been state-of-the-art for supervised learning tasks on point clouds such as classification and segmentation. In this work, a novel end-to-end deep auto-encoder is proposed to address unsupervised learning challenges on point clouds. On the encoder side, a graph-based enhancement is enforced to promote local structures on top of PointNet. Then, a novel folding-based decoder deforms a canonical 2D grid onto the underlying 3D object surface of a point cloud, achieving low reconstruction errors even for objects with delicate structures. The proposed decoder only uses about 7% parameters of a decoder with fully-connected neural networks, yet leads to a more discriminative representation that achieves higher linear SVM classification accuracy than the benchmark. In addition, the proposed decoder structure is shown, in theory, to be a generic architecture that is able to reconstruct an arbitrary point cloud from a 2D grid. Our code is available at this http URL

Advances in joint CTC-attention based end-to-end speech recognition with a deep CNN encoder and RNN-LM

Abstract: We present a state-of-the-art end-to-end Automatic Speech Recognition (ASR) model. We learn to listen and write characters with a joint Connectionist Temporal Classification (CTC) and attention-based encoder-decoder network. The encoder is a deep Convolutional Neural Network (CNN) based on the VGG network. The CTC network sits on top of the encoder and is jointly trained with the attention-based decoder. During the beam search process, we combine the CTC predictions, the attention-based decoder predictions and a separately trained LSTM language model. We achieve a 5-10\% error reduction compared to prior systems on spontaneous Japanese and Chinese speech, and our end-to-end model beats out traditional hybrid ASR systems.

Advances in Joint CTC-Attention based End-to-End Speech Recognition with a Deep CNN Encoder and RNN-LM

Abstract: We present a state-of-the-art end-to-end Automatic Speech Recognition (ASR) model. We learn to listen and write characters with a joint Connectionist Temporal Classification (CTC) and attention-based encoder-decoder network. The encoder is a deep Convolutional Neural Network (CNN) based on the VGG network. The CTC network sits on top of the encoder and is jointly trained with the attention-based decoder. During the beam search process, we combine the CTC predictions, the attention-based decoder predictions and a separately trained LSTM language model. We achieve a 5-10\% error reduction compared to prior systems on spontaneous Japanese and Chinese speech, and our end-to-end model beats out traditional hybrid ASR systems.

Deep Active Learning for Civil Infrastructure Defect Detection and Classification

Abstract: Automatic detection and classification of defects in infrastructure surface images can largely boost its maintenance efficiency. Given enough labeled images, various supervised learning methods have been investigated for this task, including decision trees and support vector machines in previous studies, and deep neural networks more recently. However, in real world applications, labels are harder to obtain than images, due to the limited labeling resources (i.e., experts). Thus we propose a deep active learning system to maximize the performance. A deep residual network is firstly designed for defect detection and classification in an image. Following our active learning strategy, this network is trained as soon as an initial batch of labeled images becomes available. It is then used to select a most informative subset of new images and query labels from experts to retrain the network. Experiments demonstrate more efficient performance improvements of our method than baselines, achieving 87.5% detection accuracy. International Workshop on Computing in Civil Engineering (IWCCE) This work may not be copied or reproduced in whole or in part for any commercial purpose. Permission to copy in whole or in part without payment of fee is granted for nonprofit educational and research purposes provided that all such whole or partial copies include the following: a notice that such copying is by permission of Mitsubishi Electric Research Laboratories, Inc.; an acknowledgment of the authors and individual contributions to the work; and all applicable portions of the copyright notice. Copying, reproduction, or republishing for any other purpose shall require a license with payment of fee to Mitsubishi Electric Research Laboratories, Inc. All rights reserved. Copyright c © Mitsubishi Electric Research Laboratories, Inc., 2017 201 Broadway, Cambridge, Massachusetts 02139 Deep Active Learning for Civil Infrastructure Defect Detection and Classification Chen Feng1, Ming-Yu Liu1, Chieh-Chi Kao2, and Teng-Yok Lee1 1Mitsubishi Electric Research Laboratories (MERL), 201 Broadway, Cambridge, MA 02139; email: {cfeng, mliu, tlee}@merl.com 2University of California, Santa Barbara; email: chiehchi.kao@gmail.com ABSTRACT Automatic detection and classification of defects in infrastructure surface images can largely boost its maintenance efficiency. Given enough labeled images, various supervised learning methods have been investigated for this task, including decision trees and support vector machines in previous studies, and deep neural networks more recently. However, in real world applications, labels are harder to obtain than images, due to the limited labeling resources (i.e., experts). Thus we propose a deep active learning system to maximize the performance. A deep residual network is firstly designed for defect detection and classification in an image. Following our active learning strategy, this network is trained as soon as an initial batch of labeled images becomes available. It is then used to select a most informative subset of new images and query labels from experts to retrain the network. Experiments demonstrate more efficient performance improvements of our method than baselines, achieving 87.5% detection accuracy.Automatic detection and classification of defects in infrastructure surface images can largely boost its maintenance efficiency. Given enough labeled images, various supervised learning methods have been investigated for this task, including decision trees and support vector machines in previous studies, and deep neural networks more recently. However, in real world applications, labels are harder to obtain than images, due to the limited labeling resources (i.e., experts). Thus we propose a deep active learning system to maximize the performance. A deep residual network is firstly designed for defect detection and classification in an image. Following our active learning strategy, this network is trained as soon as an initial batch of labeled images becomes available. It is then used to select a most informative subset of new images and query labels from experts to retrain the network. Experiments demonstrate more efficient performance improvements of our method than baselines, achieving 87.5% detection accuracy.

Joint CTC/attention decoding for end-to-end speech recognition

Abstract: End-to-end automatic speech recognition (ASR) has become a popular alternative to conventional DNN/HMM systems because it avoids the need for linguistic resources such as pronunciation dictionary, tokenization, and context-dependency trees, leading to a greatly simplified model-building process. There are two major types of end-to-end architectures for ASR: attention-based methods use an attention mechanism to perform alignment between acoustic frames and recognized symbols, and connectionist temporal classification (CTC), uses Markov assumptions to efficiently solve sequential problems by dynamic programming. This paper proposes joint decoding algorithm for end-to-end ASR with a hybrid CTC/attention architecture, which effectively utilizes both advantages in decoding. We have applied the proposed method to two ASR benchmarks (spontaneous Japanese and Mandarin Chinese), and showing the comparable performance to conventional state-of-the-art DNN/HMM ASR systems without linguistic resources.

Experimental Comparisons Between Modular Multilevel DSCC Inverters and TSBC Converters for Medium-Voltage Motor Drives

Abstract: This paper makes an intensive comparison in operating performance between a modular multilevel double-star chopper-cells (DSCC) inverter and a modular multilevel triple-star bridge-cells (TSBC) converter. Both inverter and converter are intended to drive medium-voltage motors in industrial applications. First, it makes numerical comparisons, thus, resulting in revealing that the torque and frequency of a driven motor produce a significant effect on capacitor-voltage fluctuation and arm or cluster current in the individual DSCC inverter and TSBC converter. Next, a three-phase DSCC inverter and a three-phase TSBC converter with the same rating as 400 V and 15 kW are designed and compared to drive the following two general purposes and specially-designed induction motors; one is rated at the 380-V, 15-kW, 50-Hz four-pole motor, and the other is at the 320-V, 15-kW, 38-Hz six-pole motor. This experimental comparison based on the two downscaled drive systems confirms the validity of the numerical comparison. Finally, this paper concludes that the DSCC inverter is more suitable for driving medium-voltage high-speed motors loaded with quadratic-torque-to-speed profiles like fans, blowers, pumps, and centrifugal compressors. On the other hand, the TSBC converter is more suitable for driving medium-voltage low-speed high-torque motors like mills, kilns, conveyors, and extruders.

High-Voltage Integrated Circuits: History, State of the Art, and Future Prospects

Abstract: High-voltage ICs (HVICs) are used in many applications, including ac/dc conversion, off-line LED lighting, and gate drivers for power modules. This paper describes the technologies most commonly used in commercial HVICs, including junction-isolation, thin silicon-on-insulator (SOI), and thick SOI approaches. Emerging technologies such as thin silicon membrane are also discussed.

Research and Development of the High Stable Magnetic Field ReBCO Coil System Fundamental Technology for MRI

Abstract: The superconducting magnet is effective to get a high stable and high magnetic field for magnetic resonance imaging (MRI). The current MRI superconducting magnet needed cooling in the liquid helium (4.2 K) to use NbTi superconducting wire. In the past few years, price increase and low availability of liquid helium has become a serious problem. Under such circumstances, the development of a high-temperature superconducting (HTS) coil dispensing with liquid helium cooling is greatly desired. The research and development project of the high stable magnetic field ReBCO coil system fundamental technology that started from the latter half of 2013 develops a ReBCO coil for 3 T MRI superconducting magnets. It gets a prospect of the practical use as the final aim. In this project, we will produce an HTS test coil of 300 mm bore experimentally and evaluate the magnetic field. This coil is cooled in less than 20 K by a GM refrigerator. We are going to make MRI used by the ReBCO coil field to evaluate the uniformity and stability of the magnetic field.

6.5 kV schottky-barrier-diode-embedded SiC-MOSFET for compact full-unipolar module

Abstract: For higher-voltage SiC modules, larger SBD chips are required as free-wheel diodes to suppress current conduction of the body diodes of MOSFETs, which causes bipolar degradation following the expansion of stacking faults. By embedding an SBD into each unit cell of a 6.5 kV SiC-MOSFET, we achieved, without using external SBDs, a high-voltage switching device that is free from bipolar degradation. Expansion of the active area by embedding SBDs is only 10% or less, whereas the active area of external SBDs can be over three times larger than that of the coupled MOSFET. The fabricated 6.5 kV SBD-embedded SiC-MOSFETs show sufficiently high breakdown voltages, low specific on-resistances, no bipolar degradation, and good reliability.

Wavelength- or Polarization-Selective Thermal Infrared Detectors for Multi-Color or Polarimetric Imaging Using Plasmonics and Metamaterials

Abstract: Wavelength- or polarization-selective thermal infrared (IR) detectors are promising for various novel applications such as fire detection, gas analysis, multi-color imaging, multi-channel detectors, recognition of artificial objects in a natural environment, and facial recognition. However, these functions require additional filters or polarizers, which leads to high cost and technical difficulties related to integration of many different pixels in an array format. Plasmonic metamaterial absorbers (PMAs) can impart wavelength or polarization selectivity to conventional thermal IR detectors simply by controlling the surface geometry of the absorbers to produce surface plasmon resonances at designed wavelengths or polarizations. This enables integration of many different pixels in an array format without any filters or polarizers. We review our recent advances in wavelength- and polarization-selective thermal IR sensors using PMAs for multi-color or polarimetric imaging. The absorption mechanism defined by the surface structures is discussed for three types of PMAs-periodic crystals, metal-insulator-metal and mushroom-type PMAs-to demonstrate appropriate applications. Our wavelength- or polarization-selective uncooled IR sensors using various PMAs and multi-color image sensors are then described. Finally, high-performance mushroom-type PMAs are investigated. These advanced functional thermal IR detectors with wavelength or polarization selectivity will provide great benefits for a wide range of applications.

Wide-band/angle Blazed Surfaces using Multiple Coupled Blazing Resonances.

Abstract: Blazed gratings can reflect an oblique incident wave back in the path of incidence, unlike mirrors and metal plates that only reflect specular waves. Perfect blazing (and zero specular scattering) is a type of Wood's anomaly that has been observed when a resonance condition occurs in the unit-cell of the blazed grating. Such elusive anomalies have been studied thus far as individual perfect blazing points. In this work, we present reflective blazed surfaces that, by design, have multiple coupled blazing resonances per cell. This enables an unprecedented way of tailoring the blazing operation, for widening and/or controlling of blazing bandwidth and incident angle range of operation. The surface can thus achieve blazing at multiple wavelengths, each corresponding to different incident wavenumbers. The multiple blazing resonances are combined similar to the case of coupled resonator filters, forming a blazing passband between the incident wave and the first grating order. Blazed gratings with single and multi-pole blazing passbands are fabricated and measured showing increase in the bandwidth of blazing/specular-reflection-rejection, demonstrated here at X-band for convenience. If translated to appropriate frequencies, such technique can impact various applications such as Littrow cavities and lasers, spectroscopy, radar, and frequency scanned antenna reflectors.

Vertical GaN merged PiN Schottky diode with a breakdown voltage of 2 kV

Abstract: In this study, we successfully fabricated vertical GaN merged PiN Schottky (MPS) diodes and comparatively investigated the cyclic p-GaN width (W p) dependence of their electrical characteristics, including turn-on voltage and reverse leakage current. The MPS diodes with W p of more than 6 µm can turn on at around 3 V. Increasing W p can suppress the reverse leakage current. Moreover, the vertical GaN MPS diode with the breakdown voltage of 2 kV was realized for the first time.

Implementation of continuous-variable quantum key distribution with discrete modulation

Abstract: We have developed a continuous-variable quantum key distribution (CV-QKD) system that employs discrete quadrature-amplitude modulation and homodyne detection of coherent states of light. We experimentally demonstrated automated secure key generation with a rate of 50 kbps when a quantum channel is a 10 km optical fibre. The CV-QKD system utilises a four-state and post-selection protocol and generates a secure key against the entangling cloner attack. We used a pulsed light source of 1550 nm wavelength with a repetition rate of 10 MHz. A commercially available balanced receiver is used to realise shot-noise-limited pulsed homodyne detection. We used a non-binary LDPC code for error correction (reverse reconciliation) and the Toeplitz matrix multiplication for privacy amplification. A graphical processing unit card is used to accelerate the software-based post-processing.

Integrating the Front End: A Highly Integrated RF Front End for High-SHF Wide-Band Massive MIMO in 5G

Abstract: Fifth-generation (5G) mobile communications will need to accommodate huge traffic demands in the near future. Massive multipleinput/multiple-output (MIMO) technology utilizing hundreds of antenna elements has drawn attention as a key antenna configuration for envisioned 5G applications. Realizing the massive MIMO concept of active phased-array antennas (APAAs) for 5G will require small-size, low-power-consumption, and highly accurate phase control over the wide-band frequency range, which poses significant challenges for the RF front end. This article describes prototyped highly integrated RF front ends for high super-high-frequency (SHF) wide-band massive MIMO in 5G.

A GaN PA for 4G LTE-Advanced and 5G: Meeting the Telecommunication Needs of Various Vertical Sectors Including Automobiles, Robotics, Health Care, Factory Automation, Agriculture, Education, and More

Abstract: The ever-increasing data rate and number of connections for mobile communication offer exciting user experiences in our everyday lives. Currently, the wireless communication frontier is shifting from the current fourth generation (4G) to the forthcoming fifth generation (5G). We expect much of society to go through a revolutionary change with the advent of the 5G era-a change that will involve not only the telecommunication industry but also a wide range of vertical sectors, including automobiles, robotics, health care, factory automation, agriculture, and education.

Geometric Mechanics Based Nonlinear Model Predictive Spacecraft Attitude Control with Reaction Wheels

Abstract: This paper develops a nonlinear model predictive controller for constrained attitude maneuvering of a fully actuated spacecraft with reaction wheels. In the proposed control algorithm, a Lie group ...

Image-Based Target Detection and Radial Velocity Estimation Methods for Multichannel SAR-GMTI

Abstract: In order to enhance the performance of spaceborne synthetic aperture radar–ground moving target indication (SAR-GMTI) systems, multichannel systems with large and preferably nonuniform baselines are required. In this paper, SAR-GMTI algorithms for multichannel SAR systems, which we call multichannel displaced phase center antenna (DPCA), multichannel along track interferometry (ATI), and multichannel DPCA-ATI, are presented. Multichannel DPCA is a deterministic algorithm for clutter and azimuth ambiguity suppression. It successfully suppresses not only uniform azimuth ambiguities but also nonuniform isolated ones, since it does not require uniform clutter covariance assumption as adaptive algorithms do. Multichannel ATI and multichannel DPCA-ATI are the algorithms for target radial velocity estimation. Both of them reduce the target radial velocity ambiguities, which arise with the long baseline systems, by exploiting the multiple receive channel signals. And multichannel DPCA-ATI further achieves robust performance to clutter influence by suppressing the clutter and the azimuth ambiguity in advance. The performances of the proposed algorithms are shown through airborne Ku-band three-channel SAR experiments. It is shown that the multichannel DPCA suppresses strong azimuth ambiguity up to more than 20 dB, and the accuracy of the radial velocity estimation of the multichannel DPCA-ATI is on the order of 0.1 m/s. Furthermore, statistical performance analysis is presented to discuss the potential performance on the spaceborne system.

Demonstration of 100-Gb/s/λ-Based Coherent WDM-PON System Using New AGC EDFA Based Upstream Preamplifier and Optically Superimposed AMCC Function

Abstract: In this paper, we present the first 100-Gb/s/ λ -based coherent wavelength division multiplexing passive optical network (WDM-PON) prototype system, highlighting a real-time 100-Gb/s coherent transceiver with a simplified digital signal processing suitable for access spans, and a new automatic-gain-controlled erbium-doped fiber amplifier based preamplifier with an amplified spontaneous emission compensation function for the upstream to improve the minimum receiver sensitivity for coherent detection, especially at very low received signal power. Thanks to our proposed technologies, this first demonstration achieved an increased bidirectional loss budget of more than 39.1 dB, which supports 80 km transmission with eight optical network unit splits, with an improved upstream receiver sensitivity of –38.1 dBm. In addition, to show the feasibility of the proposed 100-Gb/s/ λ -based coherent WDM-PON as a promising candidate for forthcoming 5G mobile fronthaul networks, we further investigate experimentally a 128-kb/s auxiliary management and control channel (AMCC) superimposed on the 100-Gb/s dual polarization quadrature phase-shift keying (DP-QPSK) signals employed by the 100-Gb/s/ λ -based coherent WDM-PON system. By setting the modulation index of the AMCC signals between 5% and 40%, we also successfully demonstrated 100-Gb/s DP-QPSK signal transmission with only a small power penalty of 0.2 dB due to the introduction of the AMCC signal.

Matrix Profile VII: Time Series Chains: A New Primitive for Time Series Data Mining (Best Student Paper Award)

Abstract: Since their introduction over a decade ago, time series motifs have become a fundamental tool for time series analytics, finding diverse uses in dozens of domains. In this work we introduce Time Series Chains, which are related to, but distinct from, time series motifs. Informally, time series chains are a temporally ordered set of subsequence patterns, such that each pattern is similar to the pattern that preceded it, but the first and last patterns are arbitrarily dissimilar. In the discrete space, this is similar to extracting the text chain "hit, hot, dot, dog" from a paragraph. The first and last words have nothing in common, yet they are connected by a chain of words with a small mutual difference. Time series chains can capture the evolution of systems, and help predict the future. As such, they potentially have implications for prognostics. In this work, we introduce a robust definition of time series chains, and a scalable algorithm that allows us to discover them in massive datasets.

Effect of water vapor on plasma morphology, OH and H2O2 production in He and Ar atmospheric pressure dielectric barrier discharges

Abstract: Although atmospheric pressure dielectric barrier discharges (DBDs) have a long history, the effects of water vapor on the discharge morphology and kinetics have not been studied intensively. We report a simultaneous investigation of discharge morphology, OH and H2O2 production in Ar and He DBDs operated at different water vapor concentrations and powers. The combined study allows us to assess the impact of the discharge morphology and power on the concentration dependence of the OH and H2O2 production. The morphology of the discharge is investigated by ICCD images and current–voltage waveforms. These diagnostics are complemented by broadband absorption and a colorimetric method to measure the gas temperature and the OH and H2O2 concentrations. The number of filaments in Ar DBD increases with increasing water concentration and power. The surface discharge part of the micro-discharge also reduces with increasing water concentration most likely due to a change in surface conductivity of the dielectric with changing water concentration. The OH density in the case of Ar is approximately double the OH density in He for similar power and water admixture. In contrast to the root square dependence of the OH density on the water concentration in He similar to diffuse RF discharges, the OH density in Ar increases for small water concentrations followed by a saturation and reduces for higher water concentrations. This dependence of OH density on water concentration is found to correlate with changes in discharge morphology. An analytical balance of the production and destruction mechanism of H2O2 is shown to be able to reproduce the ratio of the measured OH and H2O2 density for realistic values of electron densities.

Application of a Phasor-Only State Estimator to a Large Power System Using Real PMU Data

Abstract: This paper applies a phasor-only state estimator (PSE) to a large power system consisting of two control regions, using synchrophasor data from 56 765/345/230 kV substations. Although the technique has been previously developed, this paper describes a topology processor to determine the redundant clusters of connected buses such that the PSE can be used in real time. This PSE allows corrections for phase biases, transformer taps, and current magnitude scaling. The ability to simultaneously solve the PSE for the two control regions is due to a PMU monitoring a tie-line between these two regions. The results show that with this method, the total vector errors of the measured voltage phasor data in these two control regions average to less than 0.5% under ambient conditions. The PSE also computes virtual phasor measurements on 70 buses that do not have PMUs, including large steam generator and wind-turbine generator substations.

Experimental Verification of an Electrical Drive Fed by a Modular Multilevel TSBC Converter When the Motor Frequency Gets Closer or Equal to the Supply Frequency

Abstract: This paper provides an experimental discussion on an adjustable-speed electrical drive fed by a modular multilevel triple-star bridge-cell (TSBC) converter. The TSBC converter is suitable for medium-voltage high-power motor drives with regenerative braking. However, it suffers from capacitor-voltage fluctuation that becomes more serious as the motor frequency gets closer or equal to the supply frequency. This paper presents a practical solution of it with an acceptable increase in all the nine cluster currents. The solution is characterized by a motor-magnetizing-current controller that makes a significant contribution to theoretically eliminating the low-frequency component contained in each capacitor voltage. Experimental waveforms, which are obtained from a three-phase downscaled model rated at 400 V and 15 kW, verify exhibit satisfactory start-up performance from a standstill to the rated motor frequency that is equal to the supply frequency.

Using Volt-Second Sensing to Directly Improve Torque Accuracy and Self-Sensing at Low Speeds

Abstract: As a result of dead-time, device on-state voltage drop, dc bus voltage measurement error, etc., volt-second errors degrade precise control of torque and flux linkage, particularly at low speeds. This is true for deadbeat-direct torque and flux control (DB-DTFC), which directly manipulates the volt-second vector sourced by inverters as well as for indirect field oriented control (IFOC) drives. This paper introduces a real-time sensing scheme to measure the motor terminal volt-second vectors for each switching period with negligible phase lag. Based on the volt-second sensing, a model reference adaptive system (MRAS)-based approach is developed to decouple the volt-second errors from inverter nonlinearity and dc bus voltage fluctuation and measurement error. By delivering an accurate volt-second vector for each switching period, torque and flux control accuracy, self-sensing performance and parameter estimation accuracy are significantly enhanced.

Blockcipher-Based MACs: Beyond the Birthday Bound Without Message Length

Abstract: We present blockcipher-based MACs (Message Authentication Codes) that have beyond the birthday bound security without message length in the sense of PRF (Pseudo-Random Function) security. Achieving such security is important in constructing MACs using blockciphers with short block sizes (e.g., 64 bit).

Recent Trends in Development of High Voltage Circuit Breakers with SF6 Alternative Gases

Abstract: The available knowledge of state-of-the-art of SF6 alternative gases in switching applications was collected and evaluated in an initiative of the Current Zero Club together with CIGRE. The present contribution summarizes the main results of this activity and will also include the latest trends. The main properties and switching performance of new gases are compared to SF6. The most promising new gases are at the moment perfluoroketones and perfluoronitriles. Due to the high boiling point of these gases, in HV applications mixtures with CO2 are used. For MV insulation perfluoroketones are mixed with air, but also other combinations might be possible. The dielectric and switching performance of the mixtures, with mixing ratios that allow sufficiently low operating temperatures, is reported to be only slightly below SF6. Minor design changes or de-rating of switchgear are therefore necessary. Differences between the gas mixtures are mainly in the boiling point and the GWP.

A CW 20W Ka-band GaN high power MMIC amplifier with a gate pitch designed by using one-finger large signal models

Abstract: This paper reports a 20 W Ka-band GaN high power MMIC (Monolithic Microwave Integrated Circuit) amplifier under continuous wave (CW) operation. The one-finger large signal models were made to take account of both the phase difference of RF gate voltage at a gate feeder and thermal effect. By using this model, the gate pitch length of unit cell transistor was optimally designed to obtain maximum output power as MMIC amplifier under CW operation. As a result, 21.7W output power under CW operation was successfully achieved with power added efficiency (PAE) of 19.8% at Ka-band by a single-ended MMIC. To the best of authors' knowledge, this output power is state-of-the-art for GaN MMIC amplifiers under CW operation at Ka-band.

Power supply system

Abstract: In a power supply system using various kinds of storage batteries such as a large capacity storage battery or a small capacity storage battery, an electric power path is complicated, which may increase failures such as a short circuit trouble. There is a problem in that, when a short circuit trouble between storage batteries occurs, damage to a smaller capacity storage battery is serious. Accordingly, in a power supply system according to the present invention, the relationship among voltages, current capacities, and electric power capacities of various kinds of storage batteries such as a large capacity storage battery or a small capacity storage battery is defined so that damage to the storage batteries may be minimized even when malfunction occurs in, for example, switching a switch, and energy transfers from a higher voltage storage device.