Although internet of things (IoT) technology is widely used in various fields, IoT has several issues; energy efficiency and authentication for security. The energy harvesting technology, which harvests various energy (e.g. sunlight, vibration, electromagnetic wave, and so on), and converts them to electric power, has attracted attention for the solving method for the first problem. Moreover, the physical unclonable function (PUF), which uses variances of semiconductor manufacturing variations for authentication, has attracted attention for the solving method for the other problem. This study proposes a new authentication method for energy harvester. The proposed method utilizes the difference in power generation time due to manufacturing variations for the device authentication. In addition, the proposed method can be realized in very low cost because it does not need the dedicated PUF circuit. Several experiments using an actual energy harvester verify the validity of the proposed method.
Lightweight block ciphers, which can be implemented with low cost, are suitable for improving the safety of small devices. PRINCE is a typical lightweight block cipher and suitable for unrolled architecture implementation. In addition, PRINCE can be achieved low latency and embedded in a small area. However, it has been reported that PRINCE with unrolled architecture implementation is vulnerable to power analysis. Regarding countermeasure against power analysis, the threshold implementation and rotating sboxes masking are popular countermeasures, these require large implementation overhead though. Therefore, this study proposes a low-overhead power analysis countermeasure method for PRINCE with unrolled architecture. Moreover, evaluation experiments using a field-programmable gate array verify the effectiveness of the proposed method. In experiments, the proposed method improved tamper resistance and reduced implementation overhead in comparison with the conventional method.
We have proposed a localization method of an autonomous mobile robot using multiple sensor information in order to realize robust localization. This method is based on Particle Filter, and bias probability distributions are detected and eliminated before integration of probability distributions for localization. However, the calculation cost is enormous because it is proportional to the number of sensors and particles. In order to reduce the calculation cost for localization, we propose a new localization method using the reliability map of sensor information which is generated based on selection results of sensor information in the previous method in this paper. The usefulness of the proposed method is discussed with some experimental results with a real robot in various real environments.
The hazard perception test is one of the safety education methods aiming at the improvement of driver's attention. We have constructed a real image-type driving simulator which can extract the gaze information of the driver and dynamically evaluate the driver's hazard perception ability. In this paper, we tested the effectiveness of this system using 28 elderly drivers. The hazardous objects used for evaluation were decided based on the gaze information of driving school instructors. In this paper, we analyzed the gaze information of elderly drivers using this simulator and told the driver whether he/she was able to recognize the hazardous objects after the test was over, using the replay teaching method. After that, the driver took the test a second time. Comparing the result of the first trial (before teaching) and the second trial (after teaching), it was confirmed that the hazard perception ability improved.
For understanding the contents of a user's utterances, dialogue acts classification is often introduced to chat systems. This paper proposes Enhanced One-versus-All RNN (ENOVA RNN), which is a dialogue act classification model that consists of two RNN layers, one-versus-all layers, and the attention mechanism. The content of past utterances in dialogue is an important feature for dialogue acts classification models. In addition to the fact, classifiers tend to confuse dialogue acts that rarely appear in dialogue with frequent ones because the number of utterances differs greatly in each dialogue act. ENOVA RNN is a classification model to capture some features of dialogue acts that rarely appear after considering the content of past utterances in dialogues. In this study, it was confirmed that ENOVA RNN can classify dialogue acts using contexts not greater than six sentences. Moreover, ENOVA RNN improves rare dialogue acts classification performance keeping the overall quality of the performance by narrow down the dialogue acts using attention weights.
This paper discusses spot allocation in the national high school rugby tournament based on a quantitative strength evaluation method. In the national high school rugby tournament in Japan, each prefecture receives one spot except Hokkaido, which has two spots; Tokyo, which also has two; and Osaka, which has three. In this paper, Kanagawa and Fukuoka prefectures, whose representative teams achieved outstanding results in recent nationwide tournaments, are considered good candidates for extra spots in the tournament. This paper proposes an Elo-based quantitative skill evaluation method using about 1,800 match results. The evaluation results show high accuracy (84%) in match result prediction. In addition, the gaps between the winners and the finalists in Kanagawa and Fukuoka prefectures can be estimated. According to the results, the second team in Kanagawa ranked in the top 10 in the nationwide tournament; therefore, Kanagawa should receive an extra spot in it.
In this paper, we propose an evaluation function for the fifteen puzzle using a neural network learning random trials of moves of the puzzle. Using the IDA* algorithm with the evaluation function, we were able to solve problems of the fifteen puzzle with about one-6,600th times as less search nodes as the well-known Manhattan distance based evaluation function in average. Comparing to our evaluation function to non-admissible evaluation function whose values are products of Manhattan evaluation function and a constant between 1.3 and 1.7, the computation time were reduced to one-fifth to one-hundredth. We used data sets generated by random trials of moves from the goal pattern of the fifteen puzzle. In addition to the data sets, we chose some patterns whose distance from the goal state is 20 or 22, and which emerges frequently in the trial sequences, and we generated random sequences from the patterns. Adding those sets to the original random sequences and giving the sets to the neural network for learning, we were able to make the evaluation function more effective.
Bullying is very serious problem in Japan. To solve bullying problem as an engineering problem, our research group added some special agents into the agent-based model of bullying and verified what behavioral characteristics are most effective in suppressing bullying by using multi-agent simulation. In this paper, we introduce a new type agent (mixed type agent) who have two behavioral characteristics. We add the mixed type agent into the agent-based model and analyze the effect on bullying. As a result, a mixed type agent with relief-type and justice-type, can to reduce the number of bullying candidates.
In this study, the relationship between TBL (Team-Based Learning) performance and learning motivation was investigated. Accordingly, TBL experimental classes were conducted in university information literacy. The results revealed a moderate or weak correlation between positive scores of GAMI (Gakugeidai Academic Motivation Inventory Test) and TBL performance indicators (i.e., the pre-study time, problem-solving time, number of utterances, satisfaction level, and belief level in TBL). Furthermore, TBL performance varied depending on the number of members who had high positive scores of GAMI. The TBL performance was low for one or no individuals in a four-person formation team; however, it was high for a three-person team. These results suggest that the TBL performance of a class may improve in relation to team formation when employing positive scores of GAMI.
In this paper, we conducted a subjective experiment in order to investigate if the naturalness of vibrato of singing voice is affected by difference of vibrato waveform. In this subjective experiment, listeners judged the naturalness of vibrato of synthetic sounds with vibratos generated by human vibrato waveform and cosine waveform. It can be clarified that the naturalness of vibrato is not affected by difference of the vibrato waveform.
We propose a two-layer coding method for HDR images with noise bias compensation (NBC). Although the mean of image noise is assumed to be zero originally, it becomes nonzero (that is, noise bias is present) after tone mapping, because of the nonlinearity of tone mapping. In the conventional two-layer coding, the reconstructed LDR image is decoded using a lossy encoder of a base layer that includes noise; the noise is included in an HDR image that is inverse tone mapped from a reconstructed LDR image with noise bias. We aim to improve the quality of reconstructed HDR images by compensating for noise bias. NBC classifies pixels in a noisy image into several subsets according to the pixel value after tone mapping, and compensates for the pixel value in each subset with a compensation value determined in advance. The compensation values are determined by using all pixel values in the LDR image that is input to the base layer, in addition to those in the HDR image that is inversely tone mapped from the reconstructed LDR image. We perform NBC following inverse tone mapping for the reconstructed LDR image. Experimental results show the effectiveness of the proposed method.
Fingertip haptic feedback devices reproduce the sensation of touching objects in a virtual or remote environment to the operator, but most existing devices were proposed through a specific purpose and treat the cutaneous interaction as a point force rather than as distributed sensations. Although there are some devices which can reproduce the distribution sensations, these devices are difficult to combine the transmission system with portability very well. To solve these problems, we propose a wearable multi-point cutaneous tactile feedback device, which can reproduce different shapes and contact forces of virtual objects by using hydraulic transmission to generate physical displacement at a user's finger pad with shape-displaying method. The device consists of a tactor-array module with a widely range amplitude of 7mm used to display shapes and exert forces at the finger pad, an execution module responsible for implementing the output of the reproduction, and a circuit control module used for data processing and converting. In a virtual environment, we used a virtual spring mechanism to simulate the flexibility of the virtual finger pad and demonstrate the detection of virtual objects. An extrusion-force-deformation curve of a real finger pad was experimentally obtained and converted to allow the device to reproduce virtual contact forces at the user's finger pad. We carried out two primary interaction experiments to verify the feasibility and performance of the proposed device. The results showed that this device can display different shape patterns and reproduce the virtual contact forces at the user's finger pad.
Automatic sperm detection is in high demand for supporting Testicular Sperm Extraction (TESE). On the other hand, detection of sperms in samples of TESE is difficult because there are a lot of germ cells resembling sperms. This paper realizes automatic sperm detection for TESE by using Adaptive Thresholded Boosting (ATBoost) which is robust to overlap of feature distributions between positive samples and negative samples. In this paper, we evaluated our sperm detection method in two stages from the view point of robustness to the overlap. First, we quantitatively evaluated the overlap of the feature distributions in TESE in the metric of Bayes error rate. Second, we evaluated robustness of our sperm detection method as for the overlap. These two results show that our sperm detection method is very effective for TESE.
This paper proposes a design of decay rate controllers with guaranteed cost for Takagi-Sugeno fuzzy systems. A guaranteed cost control takes care of not only stabilization but also system performance. The controller design is based on multiple Lyapunov matrix approach, which comes from a new Lyapunov function. An illustrative example is given to show the effectiveness of the proposed controller.
In this paper, we proposed two methods to reduce the errors that happens during Electrooculogram(EOG)-based eye movement detection, especially in the case of EOG-based input system or Human-Computer-Interface (HCI). The first one is about taking measures in the detection algorithm, and the second one is by employing the Posner paradigm. Their validity was experimentally evaluated and the results show that our approaches can improve the accuracy of EOG-based eye movement detection by about 10%.
In this paper, we study operational planning and scheduling benchmark problems in an automatic picking system. These problems have been introduced as practical benchmark problems arising in logistics, and involve assignment and scheduling tasks. Simulated Annealing (SA) and Graph-based heuristics (GbH) have been proposed as methods for solving these benchmark problems. However, SA requires much calculation time and GbH does not sufficiently optimize a part of assignment tasks. We propose a new solution method combining SA and GbH for these benchmark problems. Specifically, SA determines assignment tasks and GbH determines scheduling tasks. In computational experiments for benchmark problems, we confirm that the proposed method is superior to methods using only SA or GbH in assignment and scheduling tasks. In addition, we can calculate within a realistic calculation time.
Meta-heuristics is powerful technique for acquiring a semi-optimal solution in a usable time for a large and complex system, and has many strategies such as Particle Swarm Optimization (PSO), Genetic Algorithm (GA), Ant Colony Optimization (ACO) and so on. Max-Min Ant System (MMAS) is one of improved ACO algorithms, which this algorithm sets up the range of the pheromone information for preventing overconcentration of the search area. However, MMAS has the limitation of search performance after convergence of pheromone information. In this paper, an improved MMAS using the search histories is proposed for keeping the diversity of search performance. The proposed method divides into two search groups after convergence of pheromone information, and one is the diversity group which changes new rule of pheromone update considering the search histories and the other is concentration group which uses the conventional rule of pheromone update. In the computational experiments, the proposed method has the potential to keep the search performance and diversity, and to provide the better solution.
In applying reinforcement learning to a different environment, relearning is generally required. The relearning, however, is time-consuming, and therefore a method without the relearning should be developed. This paper proposes a reinforcement learning method with generalization ability for solving an optimal routing problem with a given set of multiple goal positions. The proposed method can rapidly find the optimal route for any set of the multiple goal positions once a reinforcement learning agent learns. In the proposed method, a graph search algorithm determines the visiting order of the goal positions, and an ordinary learning algorithm such as Q-learning determines each route between goal positions. The performance of the proposed method is evaluated through numerical experiments.
Machine Learning (ML) techniques need a tremendous volume of training data. However, in operation and maintenance of industrial facilities, it is difficult to get such a volume of data due to the lack of sensors or infrequency of target phenomena. A promising approach to solve this problem is so-called data augmentation, which generates training data by using a prior knowledge or adding noise (perturbation) to original data. For this, Gaussian noise is generally used because of its simplicity. However, when the distribution of original data is not isotropic, the Gaussian-based augmentation breaks its shape, which causes so-called over regularization. In this paper, we propose a novel perturbation-induced data augmentation method, which does not require any prior knowledge and makes it easy to control the magnitude of perturbation. The novelty of the proposed method is to keep certain characteristics of shape of original distribution. The perturbation for this is generated by combined use of generative adversarial networks and newly proposed objective functions. We experimentally show that the proposed method enables to keep the gap between peaks of a mixed normal distribution. The effectiveness of the proposed method is also demonstrated in the case of an image classification task.
Knowledge distillation is a method to create a superior student by using knowledge obtained from a trained teacher neural network. Recent studies have shown that much superior students can be obtained by distilling the trained student further as a teacher. Distilling the knowledge through multiple generations, however, takes a long time for learning. In this paper, we propose a Self Distillation(SD) method which can reduce both the number of generations and learning time for knowledge distillation. In SD, the most accurate network is obtained during intra-generation learning, and it is used as a teacher of intra-generational distillation. Our experiments for image classification task demonstrate that the proposed SD acquires high accuracy with fewer generations and less learning time than the conventional method.
Maintenance tasks allow us to find faults in its early stage, and extend service life of facilities. Now, facilities and systems, which maintenance tasks are done under status monitoring, tend to be increasing, but there are still a lot of facilities that are executed under time plan maintenance. In this study, anomaly detection method is proposed with forecast error variance decomposition based on data provided from each sensor assembled in facilities.
In this letter, we report the development of a user-friendly version of the spectral sensitivity measurement device using LEDs. A large reduction in the number of circuit boards as well as board-to-board wirings has been achieved in the new architecture, making the device not only more robust but more compact than the previous version without the decrement of performance such as the variation of emission intensity.
Omnidirectional vehicles have higher mobility than four-wheel vehicles. In this paper, we study a Ball Wheeled Vehicle (BWV). This BWV is a sort of omnidirectional vehicle with three rubber balls. Three rubber balls are driven by omni wheels and the BWV can move eight directions and do pivot turn. A prototype BWV was made to single riding. But, it was a little bit large for practical use. Modifying mechanism, the BWV is downsized. Each rubber ball is controlled by a pair of omni wheel and rotary encoder with PI control law. Experimental results and comparison results between the prototype and new BWV will be given.