In this study, we have investigated model parameters of a low-field mobility model in the quantum drift diffusion (QDD) model for strained Si channel MOSFETs. Doping dependence of electron mobility in strained Si for different Ge contents in SiGe has been analyzed. The analysis has been performed by changing the Ge contents x from 0 to 0.4 in Si1-xGex. The model parameters of the mobility model are evaluated by comparing with the data calculated by using Monte Carlo method which replicates the experimental data with high accuracy. The effects on the current voltage characteristics for MOSFETs have been simulated by using the QDD model. It is found that the drain current increases as the strain increases and the rate of the increase saturates. The increase ratio of the drain current (at gate voltage is 0.7 V) is 31.0% in the range from x =0 to 0.2 and is only 4.27% in the range from x =0.2 to 0.55.
In this paper, we have investigated a quantum potential model in the quantum drift diffusion (QDD) model which allows to simulate quantum confinement effects in the inversion layer for advanced MOSFETs. The accuracy of the quantum potential model has been examined by comparing with the Schrodinger Poisson (SP) model which is able to simulate the quantum confinement effects with high accuracy. As a result, the quantum potential model shows quantum confinement effects of electron distributions in the inversion layer for an advanced MOSFET with a 1.5 nm gate oxide thickness. In addition, we have focused on the gate voltage dependence of the quantum potential coefficient in the quantum potential model. Electron densities calculated using the QDD model provide good accuracy in high gate voltage region. On the other hand, there is a discrepancy of electron densities using the QDD model comparing the SP model in subthreshold voltage region. It is found that the discrepancy comes from an assumption of constant model coefficient.
Molecular orientation, molecular packing structure, and electrical properties of thermally stable phthalocyanine, tert-butylated phthalocyanine (t-BuPcH2) in the thin film fabricated by bar-coating method were investigated. The XRD patterns indicated that the columns composed of t-BuPcH2 molecules formed hexagonal structure, the axis of which was parallel to the bar-sweep direction, and t-BuPcH2 molecules were vertically aligned to the bar-sweep direction. Uniaxially oriented thin films of t-BuPcH2 were successfully fabricated on Si substrate covered with polyimide by bar-coating method, and the field effect mobility of t-BuPcH2 was estimated to be 1.32×10-4 cm2 V-1 s-1 at maximum. The field effect mobility parallel to the bar-sweep direction was approximately three times higher than that perpendicular to the bar-sweep direction due to high electrical conduction along the column axis of t-BuPcH2.
Instabilities due to mutual interactions among inverters and grid systems are sometimes observed. This paper proposes an optimal digital controller design method for passive stabilization of a grid-connected three-phase inverter with LCL filter to improve output frequency impedance characteristics which sufficiently satisfy the Nyquist stability criterion. The principle is that the output impedance can be designed to be passive with the designed feedforward loop characteristics and the system is assured to be stable even if the grid system is weak. The proposed three-phase control consists of the following three steps. In the first step, the state feedback control gains for a single-phase control subsystem are designed to have a good sinusoidal tracking characteristic according to the optimal control. The designed characteristic of the output impedance may not be passive at a certain frequency band. In the second step, a resistive characteristic is added to the band by a new feedforward control design based on a fitting method to make the total characteristic passive and to ensure stability. In the third step, combining two sets of the designed single-phase control subsystems, a three-phase control system is derived. The principle and the design process are described and the designed system is investigated and validated for stability and response characteristics through simulation and experimental results.
Recently, various technologies using PointCloud and Deep Neural Network (DNN) have been actively researched. However, there is a disadvantage that the collecting PointCloud data from real object with special sensors such as depth sensor is time consuming task. To deal with this problem, we focus on 3D reconstruction from a single image. Conventional methods construct PointCloud from a single image which includes mask information. Therefore, it is difficult to construct a PointCloud from an image without mask information. To remove the requirement of the additional information such as mask for input image, we propose data augmentation based on style transfer for 3D reconstruction. It is known that DNN using style transformed image can learn a shape feature. By using the transformed images, the DNN can learn object shapes with various backgrounds and textures and can obtain shape features even from the images with cluttered background. From the experimental results, we confirmed that our proposed method could construct 3D object shape with PointCloud while keeping shape information without additional information.
Recently, speech synthesis has been spotlighted as a key technology for broadcasting original movie with character on YouTube. To make a natural speech in the methods based on GAN(Generative Adversarial Network), the following unsolved problems are remained: impression of synthesized speech such as warm, cool, etc., and long-term optimization of speech synthesis. In the former problem, since the conventional methods have focused on natural intonation of speech, they have not discussed the impression sufficiently. In this research, to deal with the impression, we proposed a new GAN based speech synthesis method using impression vector digitized the speaker impression. On the other hand, for the latter problem, since conventional methods optimize the relationship among frames insufficiently, the synthesized speech is still not natural. To solve this problem, inspired by an image synthesis technology such as HDGAN, we proposed a new GAN based network structure. The characteristic point is hierarchically nested discriminators at intermediate layers of the generator. In experiments with 15 speeches synthesized by the proposed method and 14 impression items, we estimated impression recognition accuracy by 11 listeners as subjective evaluation. From the experimental results, we have achieved 40.61% of subjective accuracy.
Recently, layer stack approach for CNN (Convolutional Neural Network) has achieved high image recognition performance. However, as the number of stacked layers is increased, this leads to increasing number of parameters. Therefore, a high-speck machine is required for calculation. To solve this problem, in this research, we focus on SE module, which has an attention mechanism that adaptively trains the relationship among filters and achieves the improvement of recognition performance with a slight increase in the number of parameters. Although this module has achieved high parameter efficiency compared with the layer stack approach, the adaptability of the SE module is discussed insufficiently. Therefore, in this paper, we evaluate its adaptability by utilizing DenseNet and ResDenseNet, which have higher parameter efficiency compared with ResNet and both have DenseBlock modules. Unfortunately, a simple combination of SE module and such CNNs generally adds the SE module to end of each CNN module, which leads increasing a large number of parameters since the SE module requires parameters depending on the number of filters in DenseBlock. To solve this problem, we propose a new combination of SE and DenseBlock modules, that is, we add the SE module to each branching function. From the empirical evaluation with CIFAR10 and CIFAR100, our proposed method improved recognition performance compared with DenseNet/ResDenseNet without SE modules.
This paper describes optimization of a proposed high reliability and wide SOA 100 V N-LDMOS transistor for automotive applications. The drift region of the device is enclosed with two P-type buried layers, dual RESURF structure, and the field plate forming a two-step structure is grounded. The drift region and the field plate were optimized to obtain high hot carrier endurance, high suppression of drain current expansion CE, high breakdown voltage BVDS, and low specific on-resistance RON,SP taking mass production into account. Within the mass production tolerance, the electric field near the gate-side drift region edge of the proposed device is about 70% of that of a conventional device under a high hot carrier generation condition, the drain voltage causing CE of the proposed device is about 20 V higher than that of the conventional device under a high CE generation condition, and the BVDS - RON,SP characteristic of the proposed device is at a state-of-the-art level: BVDS = 131 V and RON,SP = 150 mΩmm2 at the worst case. Furthermore, due to slight changing the tolerance range, high ESD endurance of the proposed device could be obtained at the expense of RON,SP and suppression of CE a little bit.
Information recommendation using UGC (User Generated Contents) systems has been paid much attention recently. Explainability and reliability are important for users easy to interpret the search results of a recommendation system. This paper proposes a hierarchical concept network that represents inferences based on evaluation criteria for recommendation in order to satisfy explainable recommendation. Futhermore, we also achieve reliable recommendation by restricting knowledge sources in generating the links between concept categories. We evaluate that our proposal method can generate explainable and reliable search results in some examples of restaurant information recommendation.
In the field of commercial buildings where energy conservation is strongly promoted, a building energy management system (BEMS) is concerned by the energy managers of the buildings. However, the energy manager cannot overcome the problem called “energy-saving barrier” just by introducing conventional BEMS and visualizing energy consumption. In order to solve this problem, this paper proposes an energy-saving service which utilizes the large amounts of data from the BEMS. The service can suggest some improvements of the operation as well as indicate the return on investment of replacement of facility. The improvements of the operation are almost automatically suggested by using the singularity detection method. The service is applied to the commercial buildings including large, medium and small size. The 5-year experimental results show the usefulness of the proposed service and method.
In the actual sound environment system, a specific signal shows various types of probability distribution, and the observation data are contaminated by an external noise (e.g., background noise) of non-Gaussian distribution type. Furthermore, there potentially exist various nonlinear correlations in addition to the linear correlation between the system input and output time series. Consequently, the relationship between the system input and output cannot be represented by a simple linear model based on only the linear correlation and lower order statistics. In this paper, a complex sound environment system difficult to analyze by using usual structural method is considered. By introducing a nonlinear system model based on conditional probability distribution with various correlations between input and output signals, a prediction method of output response probability for sound environment systems is theoretically proposed in a suitable form for the system with nonlinear, non-Gaussian and nonstationary properties. The effectiveness of the proposed method is experimentally confirmed by applying it to the observed data in sound environment.
In this paper, portfolio optimization using loan is formulated as a chance constrained problem in which the money borrowed from a loan is invested in risk assets. The chance constrained problem is transformed into an equivalence problem and proven to be a convex optimization problem. For deciding a proper interest rate of the loan that benefits both borrowers and lenders, a new method is proposed. It is shown, both theoretically and empirically, that the loan is always used up to the limit for improving the efficient frontier if the interest rate is decided by the proposed method.
Measurements performed using the self-coupling effect of a semiconductor laser utilize an external resonator by exploiting the reflected light from the surface of an object. The improvement of the measurement accuracy due to the use of self-coupled signals acquired from terminal voltages has not been thoroughly investigated. The current that flows through a semiconductor laser is adjusted so that its wavelength changes at a specific rate when it is modulated using an arbitrary waveform. As such, the spread of the self-coupled signal is suppressed, and precision is improved. In addition, the accuracy is improved by considering the influence of lenses on the optical path.
In recent years, super-resolution using deep learning has attracted attention. Super-resolution is a technology for converting low-quality images to high-quality images. Super-resolution can be applied to the technology to identify a criminal from the video of a security camera. It is difficult to identify the criminal from raw images, because security cameras are low image quality to record long-term images. In this study, we propose a method to super-resolution human face images using Capsule network. Capsule Network represents input values and output values as vectors, which makes it possible to learn features between a positional relationship and an orientation of faces. Therefore, it can be expected to generate a face image of higher quality than Convolutional Neural Network (CNN). We employ the CelebA data set,which is collected about 200,000 face images, as the training data. The low quality image is generated from the original CelebA image. Capsule network is trained using original high quality images as outputs and low quality images as inputs. Experimental results show that the super-resolution method using capsule network generates high quality face image than CNN.
In order for local government and police authority to effectively prevent local crimes, under financial and human constraints surrounding the local government, it is important to appropriately share roles and collaborate with local residents, after understanding whether or not the “local eyes” by residents works. However, since the “local eyes” are intangible attitudes and actions of residents, it is difficult to estimate them quantitatively. In this study, based on the understanding of that the behavior of reporting by citizen report system represents a part of the function of “local eyes” in a region, the relationship between the citizen report rate via citizen report system and the crime rate were analyzed. Then, it aimed to clarify that it is possible to estimate whether or not the “local eyes” actually functions by analyzing the rate of citizen report. In conclusion, this study revealed that in urban areas and commercial areas, there is a correlation between the citizen report rate and the crime rate, and it is possible to estimate whether the “local eye” works.
Recently, factories which produce high-mix assembly products face challenges to estimate the planned man-hours accurately and automatically, for the purpose of reducing overtime hours. In factories which produce high-mix assembly products, it is difficult to estimate the planned man-hours because the number of products is small and there are few opportunities to produce similar products. In order to solve the above problems, this paper proposes an approach with the following two features. The first feature is to calculate the man-hours for each product specification by the regression analysis with the appropriate upper and lower limit values set as the partial regression coefficient as input for the actual man-hours data and product specification data for each product / process. The second feature is to regularize with an appropriate value as product specification in order to stabilize the solution even when the amount of data is small. We have applied the approach to our factory, and we proved the approach can reduce 18% of the rate of deviation between the planned man-hours and actual man-hours at a maximum compared to the case of planning experts. Therefore we have confirmed the effectiveness of the proposed system and released as a product.
In this paper, a fast convergence adaptive filter with square sum of the correlation function as a cost function is studied. Since the algorithm does not include the inverse correlation matrix of the input signal of the adaptive filter, it is stable and has only a few arithmetic quantities. The convergence performance of the algorithm is also verified.