It is an important issue to protect the intellectual property of software. The authors proposed to conceal some part of software by implementing it as logic circuit. Though the security is further improved by obfuscating the logic circuit, it requires much effort to develop the dedicated obfuscation tool. Matsuoka et al. proposed using the software obfuscation tool, Obfuscator-LLVM (oLLVM), with C-backend (CBE) and Xilinx Vivado HLS (high-level synthesis) to generate the obfuscated logic circuit. This study proposes adopting another HLS, LegUp, to obfuscate logic circuit. The feasibility of these two methods are examined with 12 applications of CHStone benchmark, each of which were evaluated with four obfuscation methods (total 48 cases). In our experiments, Matsuoka's method failed to generate the correct hardware in 5 cases out of 48 cases. Meanwhile, the proposed method successfully generated the correct hardware for all 48 cases. The average latency derived by the proposed method was 43% larger than that by Matsuoka's method. The logic scale derived by the proposed method was also 42% larger (LUT) and 112% larger (FF) than that by Matsuoka's method.

Asynchronous circuits are a technique in order to solve the some serious synchronous circuits' problems such as clock skew, power consumption and electromagnetic noise. In this paper, we designed histerisial variable-threshold gates based on neuron MOS transistors for asynchronous circuits. The proposed hysteresial variable-threshold gates are also expected usuful for neural networks or machine larning for the hysteresis, that is memory, characteristics. The developed gates having variable-threshold operations can be used to set a threshold of a neuron. The simulation results of hysteresial variable-threshold gates are provided with SPICE simulator. The synthesized hysterisial variable-threshold gates has three gate-inputs and two control wires. Then asynchronous half-adder is demonstrated as an arithmetic circuit example.

An alternative learning and its application to construct a robust object recognition system for intelligent security camera(RORSIS) is presented in this paper. We show some experimental results of the development of the new robust 3-D object recognition system for intelligent security camera. In this system, a deep neural network and alternative learning using 3-D CG are key techniques for object recognition from a free viewpoint. Alternative learning is effective approach for the machine learning that depends on huge amount of tarining data. Some appearance based characteristics are determined from captured images, and the system uses a deep neural network, called AlexNet, to automatically classify bicycle, automobile and so on. The proposed system shows that several kinds of equipments can be recognized from a free view point. Experimental results show that the system can effectively recognize four kinds of real objects with 99.5% accuracy.

This paper describes the development of 3-D human recognition system of flying drone system for emergency rescue and investigation. In this system, deep neural network and its application for 3-D object recognition are key techniques for human detection from a free viewpoint. Some appearance based characteristics are captured as movie frames, and the system uses deep neural networks to automatically classify concerned object. The proposed system performs well that many kinds of views of personnels can be recognized from bird's-eye view. Experimental results show that the system can effectively recognize human objects with high accuracy.

Achieving practical and full-scale use of drones will require a transition from first-person-view (FPV) flight based on visual radio control to autonomous wide-area, long-distance flight. However, technology that enables drones to fly autonomously over a wide area and long distances while feeding back positioning in ever-changing real-world environments has yet to be established. It is essential that R&D efforts take up both waypoint flight using GPS data and non-GPS autonomous flight. We have previously developed an integrated drone navigation system that combines autonomous waypoint-GPS flight and non-GPS autonomous flight using positioning based on ground images taken by an on-board camera. We have since been developing an autonomous wide-area, long-distance flight system that feeds back drone positioning in actual ever-changing environments via the public cellular network using an on-board mobile phone. We have achieved good results in basic operation trials for both systems demonstrating their potential for practical use.

In this paper, we propose a new intelligent security camera system, that is named COMDES, for automated detection of snatching incidents on streets during the night by using LSTM network. Although over a half of all snatching incidents occur at night, this has not been considered in past studies. Thus, we have proposed an intelligent security camera system using a deep neural network and snapshot of a video frame to detect snatching incidents in the night by our previous paper. The COMDES can perform more efficient detection of snatching than our previous paper, by using sequential frames observed in the criminal scene and LSTM. It can classifies the situations into criminal or non-criminal scenes precisely. The experimental results show that the system, COMDES, can effectively detect snatching incidents with an accuracy of 98.89%.

All of the peaches exported to Taiwan are obliged to inspect by the quarantine regulations in Japan. In our research, the system which inspects the damage of the peach by the larva of peach fruit moth, Carposina sasakii Matsumura, using an X-ray penetration system is researched and developed for the purpose of export promotion. This paper describes the components of the inspection system of insect damaged peaches, and the inspection method using X-ray penetration images. The verification tests of our system were done in 2016, 2017 and 2018. The subjects which became clear in 2016, 2017 and the solutions to there are described. The results are reported, which are the results of the verification tests by the improved system in 2018.

The classical music “uta-sanshin” has been sung since the Ryukyu Kingdom period, and its skills commonly depend on folklore method by bush telegraph. Accordingly, there exist much sensibilities and esoteric expressions of the uta-sanshin expert in passing down the skill. Also, the decrease in number of successors accompanying aging and the difficulty in understanding the musical score are hindering the inheritance and the reconstruction of the music. In this paper, we apply the deep learning to Ryukyuan classical music and develop a system that identifies vocalism by real-time processing. The results of the evaluation, compared with the conventional method, show that the execution time is reduced to 98%, and the identification accuracy is improved by 6%.

Recent camera can take a bright picture even in dark field. The Processor of camera gain signal. However noise is amplified when signal amplified. In this article, we propose an easily technique bilateral filter for noise reduction. The proposed technique can efficient noise reduction in 4 times of filtering and the optimized parameter searching time is drastically shortened.

Potentials of lateral hetero superjunction (HSJ) power devices, that are based on III-V semiconductors such as GaN and GaAs, are analyzed theoretically to clarify how the product R_ON·A of their on-resistance R_ON and device area A depends on the breakdown voltage. By focusing on power devices having the breakdown voltage of about 1kV, we clarify to which level the specific on-resistance R_ON·A of various devices can be reduced by the selection of the device structures and materials. It has been found out that the lowest limit of R_ON·A for HSJ devices can be lower than that of Si insulated gate bipolar transistors, and also than that of vertical power FETs based on SiC and/or GaN. We note in particular that not only GaN-based HSJ devices but also GaAs-based HSJ devices possess outstanding potential because of the high electron mobility nature of these heterojunction structures.

In this study, the movement strategies of the sit-to-walk clarified from the phase plane trajectory of the body center of gravity based on posture stability and the equilibrium-point and stiffness analysis based on muscle synergy and equilibrium-point hypothesis. 10 healthy young subject performed sit-to-stand at slow, natural, and high speeds. In addition, subjects performed sit-to-stand-and-walk and sit-to-walk at a natural speed. Then, the mean value of the root mean square values was calculated for the equilibrium-point trajectory and stiffness, along with and the angle and angular velocity in the argument angle direction. Results revealed that the stiffness adjustment is more influenced by in the stability of the sit-to-walk than the equilibrium-point trajectory. Also, it is necessary to increase the stiffness in the argument angle direction compared to the sit-to-stand at slow. In the phase-plane trajectory, the angular velocity considerably influences the posture stability than compared with the angle. Some subjects performed sit-to-walk with applying a strategy similar to that applied for the sit-to-stand-and-walk mechanism. Thus, the sit-to-walk can be performed without increasing the stiffness in the argument angle direction.

Epilepsy is a neurological disorder characterized by abnormal electrical activity of neurons affecting a small or large area of the brain. Epilepsy is categorized into focal epilepsy and generalized epilepsy. Focal epilepsy is due to abnormal electrical activity that effects the smaller brain sections, whereas in generalized epilepsy the larger section of the brain is affected. Epilepsy affects the patient's personality, behavior, and emotions, making daily routines difficult. Early detection is therefore essential. In this study, we use electroencephalogram (EEG) signals to detect the focal epilepsy. However, due to the random and non-stationary characteristics it is difficult to analyze the subtle changes in the EEG by visual inspection. Hence, in this study we proposed a system that classifies focal epileptic and normal signals accurately. In order to analyze the difference in the randomness of the focal epileptic and the normal signals we have used fuzzy approximation entropy and spectral entropy. The system first decomposes the signal, and then features are extracted using fuzzy approximation entropy and spectral entropy. We obtain classification accuracies of 95% and 85% using 4-nearest neighbor and support vector machine classifiers, respectively.

This research proposes a new service value evaluation model for social infrastructure service. Existing service concept or model is designed based on the direct interaction between providers and customers. By the direct interaction, providers know the customers' need, then they provide customized and delicate service to each customer. Therefore, the service level becomes higher and customers' satisfaction and loyalty glows. In contract, we pay attention to social infrastructure service. In social infrastructure service, there are few direct interactions between providers and customers. In addition, providers cannot provide the customized service to each customer although they have variety of needs. For social infrastructure services which have different characteristics, we propose a new service concept “SISLA (Social Infrastructure Service Level Agreement)” based on the concept of SLA (Service Level Agreement) and a new service value evaluation model in this paper. By this model, we show the possibility that social infrastructure service providers may be able to make an appropriate judgement to provide more better service for customers while ensuring profits.

In order to enable an autonomous mobile robot to travel over rough terrain, it necessitates the capability to detect self-position accurately even when the odometry errors are increased in traveling. The conventional method can keep high speed and precise localization using ICP algorithms or feature matching techniques. However, effects of steep changes of a mobile robot position are not considered when it travels over rough terrain. In this paper, we propose the method for efficient real-time 6D pose tracking using a rotating 2D laser scanner in traveling over rough terrain. For adaptation to steep changes of the position, weighted pointclouds are generated based on the angular and the linear velocity measured by sensors mounted on the robot. And the position and posture of the robot are sequentially optimized by the scan matching in increments of 10 scans. In indoor experiments, we evaluated accuracy of our method when the robot passes on rugged floor. As a result, our method was performed less than 0.078m RMS positional error in real time.

Robots that have intimate contacts with people such as communication robots and care assistant robots are used widely in recent years. Appearances of such robots are important to encourage use of the robots and to extract desirable reaction from people. Also, each demand for such robots contains minor differences and each robot for the demand is designed one by one. However, designing robot spends long time and each demand would be changing every day. For these reasons, designing robots needs much more speed and capability of realizing desired appearance. In this paper, for automatic robot designing with arbitrary shape and functions by a computer, a uniform method to simplify robot designing by 3D printed inner structure is proposed. Robot designing is converted into optimization problem of allocating components by proposed method and becomes suitable for a computer to deal with. Desired shape is also maintained while needed functions are equipped by this method. Effects of proposed method is confirmed through shape comparison experiment.

The structure and weight in Block-Based Neural Network (BBNN) are optimized by utilizing genetic algorithm. The architecture of BBBN consists of a two-dimensional (2-D) array of basic block with four input/output nodes and connection weights for block's output. To propose easier hardware implementation like Field Programmable Gate Array (FPGA), integer weights are used in the basic block. Each block can be one of the four different basic types and the architecture of BBNN is configured with the combination of basic block internally configured. However, BBNN's structural change needs hardware reconfiguration and the cost is very high. To reduce the reconfiguration cost, Smart Block-based Neuron (SBbN) which has sufficient number of weights for all four types of basic block has been proposed. SBbN preserves all weights even unnecessary for some types, and thus it consumes redundant hardware resource. A new model of BBNNs in which all weights in SBbN are used efficiently with modifying calculation procedures of outputs of basic blocks has been proposed and it eliminates the resource redundancy of SBbN. However, new approach which both, left and right's side nodes concurrently serve as input and output does not provide parallel computation in left and rightward signal flow. This paper presents a parallel computation with independent side nodes for each signal flow.

Recently, convolutional neural networks (CNNs) have achieved remarkable results on sentence classification problems. In these approaches, each word in the sentences is transformed to real number vectors (called word vectors) and the sentences as input data to the CNN are represented by the sequences of the word vectors. A dataset for training and testing for the CNN includes the large number of words, therefore the word vectors are embedded so high-dimentional space. As a result of this, the input data space of the CNN becomes very high. When the input data have high dimension, much training data are required for enough training of the CNN. It is not always possible, however, to get enough number of data for training. If the enough data cannot prepare for learning, it is desirable to decrease the dimension of input data. This paper shows the results that the smaller dimensional word vectors are applied to sentence classifications by CNNs. The results have shown that some dimensionality reduction does not effect too much to the accuracy of the sentence classifications by CNNs.

The analytic hierarchy process (AHP) is an effective method for product selection support because it evaluates alternatives based on the total weight, which is a quantitative value. The total weight is obtained by multiplying the ratio of each evaluation criterion for the target user and the weights of alternatives in each evaluation criterion. The determination of weights in the qualitative evaluation criterion is difficult because of the necessity to compare the questionnaire results of all alternatives.

This paper proposes Decision Support System for Product Selection based on AHP, using the decision rule of Rough Set for Qualitative Evaluation. The proposed system makes decision rules based on the target user's judgement of “Good” or “Bad” on several samples in the qualitative evaluation criterion. Decision rules have a qualitative and quantitative evaluation value for each attribute, and the system calculates the weight of the AHP through normalization of these evaluation values. The user can know the results easily, because the evaluation method of alternative samples in the proposed system is the same as that of correspondence with the shop assistant in an actual store. Based on two experiments, we confirm that the setup of attributes for decision rules of rough sets is one of the most important elements of the proposed system.

Servitization of manufacturing industry has become important problem. However, there are very few studies which show specific methods in the manufacturing industry. In this paper, we propose a knowledge whitespace model concept as a value co-creation framework with customers based on Service Dominant Logic in the manufacturing industry. Separate the functions of the target service / product into two hierarchies, the application layer and the platform layer, and remove the functions other than the most important function. Therefore, this model intentionally creates knowledge whitespace and promotes value co-creation with customer. We analyzed concrete examples of development of multiple services and products and examined the effectiveness of the model focusing on relationships, information exchanged, co-created value.

This paper presents a compact feed circuit using microwave switch which controls beam directions. The feed circuit consists of two hybrid circuits, transmission lines and three switches, and it has one input and four output ports. The switches consist of the lumped elements and PIN diodes, and they are set to three states to change phase differences between antenna elements (namely 0°, 90°, or -90°). The feed circuit is designed at 1.0GHz using microstrip lines, and characteristics of the circuit are evaluated. The measured results almost coincided with simulation.

In this paper, the estimation accuracy of a second order cascaded IIR adaptive notch filter using adaptive line enhancer is studied. It is shown that the bias error of the convergence value of the tap gain from the optimum value can be greatly reduced by using the output signal of the line enhancer as the reference signal in the vicinity of the solution. The bias error is formulated and the comparison is made by the theoretical and the numerical verification for the case when the adaptive line enhancer is not used.

The purpose of this research is to confirm that moving object detection is possible from surveillance camera image using change of DCT signs. First, change of DCT signs are detected between the current frame and the background image. Then, only the changed DCT signs were inversely transformed and binarized. As a result, it was possible to detect the edge part of the moving object. In conclusion, it was found that moving object detection is possible from the change of DCT signs of the surveillance camera image.