Recently, gallium arsenide (GaAs)-based resonant tunneling diode (RTD) oscillators and indium phosphorus (InP)-based Schottky barrier diode (SBD) receivers have been studied in the terahertz (THz) band. The THz devices for practical use should operate at room temperature, be small in size, and have high output power. Therefore, this study was focused on gallium nitride (GaN), which possesses excellent material properties, such as wide bandgap characteristics and heteroepitaxy on Si substrates. The GaN-based oscillators and receivers are expected to be compact, operate at room-temperature, and act as a high-power device for the THz-band devices. However, GaN has crystal defects, which can cause instability in device operations. A double dielectric structure patch antenna composed of Silicon Nitride (SiN) and Benzo Cyclo Butene (BCB) with different dielectric constants was proposed to realize a GaN-based THz transmitter and receiver. The antenna characteristics were investigated using the Finite Difference Time Domain (FDTD) method. The results showed that the SiN has little effect on the radiation, whereas the BCB is strongly responsible for the radiation. Comparing the absolute gain between the double dielectric structure and the conventional structure using the SiN, it was confirmed that the double dielectric structure can improve the absolute gain.
Efficient computation of eigenvalues and eigenvectors of the Gram matrix for quantum signals is desirable in the field of quantum communication theory. Because various quantities such as the error probability, mutual information, channel capacity, and upper and lower bounds of the reliability function can be obtained by the eigenvalues and eigenvectors. Moreover, solving the eigenvalue problem also provides a matrix representation of quantum signals, which is useful for simulating quantum systems. In the case of symmetric signals, analytic solutions to the eigenvalue problem of the Gram matrix have been obtained and efficient computations are possible. However, for asymmetric signals, there is no analytic solution and universal numerical algorithms must be used. Recently, we have shown that, for certain asymmetric signals, the Gram matrix eigenvalue problem can be simplified and it is sufficient to consider the eigenvalue problem of smaller matrices at half or quarter size. In this paper, we consider nm-ary quantum signals formed by rotating signal points in a circular sector region and show that solving the 1/n-size matrix eigenvalue problem is sufficient to give the eigenvalues and eigenvectors of the Gram matrix.
In Japan's super-aged society, it is important to prolong independent daily living. This paper, we focus on walking, which is indispensable for daily living, to diagnose frailty. This study included acceleration data during limb-loaded walking to increase muscle strength in the analysis. Also, we classified and identified whether middle-aged and elderly persons are in a frail state or not were conducted following the Cardiovascular Health Study. After selecting useful features by random forest, a Support Vector Machine was used to identify the frailty state. As a result, we reported that we could identify frailty with an accuracy rate of 82%, suggesting the possibility of detecting low grip strength from the features extracted from gait data.
According to earlier studies of distance-velocity sensors using the self-coupling effect of semiconductor lasers, the linearity of the triangular wave current that drives the semiconductor laser was poor. Therefore, the linearity of the triangular wave was improved using arbitrary waveforms to improve the accuracy of the measured values. However, multiple triangular waves were required for measurement; hence, the measurement was time consuming. Therefore, we changed the driving circuit of the semiconductor laser to a current spitting method to improve the linearity of the triangular wave current and conducted research on signal processing for mode-hop pulses, which are measured with a single pulse. Consequently, we devised a new method that enables single-pulse measurement while maintaining the measurement accuracy of the sensor. The experimental results indicated that distance and velocity can be measured with the same level of accuracy as the results obtained with multiple pulses to date.
Information on fabric material is necessary in washing and ironing clothing. However, indication on a care tag may peel off or the tag may come off due to deterioration over time. Discrimination of the material from the fabric itself is not easy for a general person. Estimating the material of an object is one of the challenging tasks in computer vision. This paper deals with the identification of cloth materials using computer vision. We studied a method to discriminate the fabric material from the image of clothing taken by a smartphone camera. First, we investigated the relationship between image resolution and discrimination accuracy using a convolutional neural network (CNN). As a result, we observed that the accuracy changes with resolution and that the resolution at which the accuracy is highest differs depending on the material. Based on these results, we proposed a fabric material discrimination method using multi-resolution images by combining two CNNs. As a result of the evaluation experiment, the proposed method discriminated six kinds of fabric materials with 87.1% accuracy, and the accuracy was significantly higher than that of the comparison method without using multi-resolution images.
We propose methods of panel extraction and plane background estimation for motion comic creation of a manga image. In the panel extraction, the speech balloon extraction method proposed by the authors is utilized for sharpening a dividing line of the panels. Frequency of use of background techniques is investigated, and it indicates that the frequency of the plane background is high. The targets of the plane background techniques are blank space, solid, and screen tone, and the plane backgrounds is estimated in this paper. In the experiment, recall values of the panel extraction and plane background estimation were 88.6% and 93.5%.
With the increasing importance of cyberspace, the threat of malware has also been increasing. In order to establish anti-malware technology, it is important to clarify the threat of malware. Therefore, this study considers new malware using steganography. Evaluation of this study demonstrates that conventional security software cannot detect the proposed method which embeds the malicious code in images. Furthermore, this study proposes a new detection method which utilizes bit-plane in image file with malicious code to overcome the malware. Experiments verified the validity of the proposed method. It was also shown that the proposed detection method is also effective for variants with altered payloads.
It has been theoretically predicted that the skeletal muscles show peculiar frequency-dependence of the electrical impedance between 1 Hz and 100 Hz; the resistance increases with frequency and the reactance becomes positive. In such low frequency-region, experimental results have not been reported, and instruments for use in the bioelectrical impedance analysis are not commercially available, at this stage. In this study, measurements of the impedance of a test circuit consisting of capacitors and resistors were carried out by the four-terminal method using an instrument for electrochemical measurements (PARSTAT 3000A, Princeton Applied Research) in a frequency-region from 0.1 Hz to 1 MHz in galvanostatic mode in which the amplitudes of the applied current were 100 µA (rms) and below, that met the safety standard JIS T0601-1. When the amplitudes of the applied current were between 100 and 30 µA (rms), observed values of resistance and reactance were in good agreement with theoretical ones from 0.1 Hz to 100 kHz. The capacitance and conductance agreed with theoretical ones below 30 kHz and below 3 kHz, respectively. These results suggest that the instrument is available for the bioelectrical impedance analysis down to the low-frequency region under the condition required by the safety standard.
The process tomography (PT) method using ultrasonic waves or electromagnetic waves has been proposed to measure the bubble distribution in a pipe where a liquid containing bubbles flows, but it has a problem of making the equipment larger and more complicated. In this paper, we propose a simpler method for imaging the distribution of air bubbles in a pipe by using an M-sequence coded array probe as the ultrasonic transmitter and receiver. In this paper, the effectiveness of the proposed method was verified by simulation using the finite difference time domain (FDTD) method and actual experiments. As a result, it was found that it was possible to image objects in a circular area in water in both simulation and actual experiments, suggesting that this method can be used to instantly image objects flowing in a pipe.
This paper proposes a persistent coverage control method on a sphere, taking robot safety into account based on control barrier functions. First, we present a persistent coverage control on a sphere, which dynamically changes the arrangement of multiple robots with a limited-range sensor. Second, we also consider the safety of the robots on the sphere by using control barrier functions that guarantee energy management and collision avoidance. Furthermore, the safety-aware persistent coverage control is reduced to a quadratic programming problem using the control barrier functions, and optimization-based control law is proposed. Finally, the effectiveness of the proposed control method is demonstrated by simulation.
For realizing desired control performances, data-driven tunings such as virtual reference feedback tuning (VRFT) and fictious reference iterative tuning (FRIT) were proposed. Furthermore, estimated response iterative tuning (ERIT) can realize desired closed-loop response of two-degree of freedom controls. However, conventional data-driven control cannot realize desired control performances of autonomous vehicles because vehicle dynamics is time variant systems. To solve this problem, we propose time-varying feedforward controls based on ERIT. The validity of proposed method is verified through experiment verification.
In recent years, the number of websites has been increasing, and with it, the number of similar services. As a result, UX is becoming more and more important to measure the differentiation of a website from its competitors, and UI plays a major role in improving UX. Website developers implement a development process called HCD to improve UI. However, in HCD, there is a gap between the perception of the evaluator and the developer, resulting in a lot of man-hours. In this paper, we propose a method for evaluating web pages using GUI operations to solve this problem. We let the evaluator of the web page edit the web page by mouse operation from the browser, and feedback the result directly to the developer. As a result of the experiment, we were able to discover problems with the web pages that could not be obtained by the conventional questionnaire. It also made it easier to design improvement plans, leading to a reduction in man-hours.
To find lung diseases, physicians need to conduct various examinations. Recently, to reduce their burdens, many applications of deep learning have been proposed to diagnose chest X-ray images. However, there are few studies using deep learning for auscultation, and also, there are only a few small-scale benchmark datasets of lung sounds that are annotated for machine learning. Therefore, we aim to build an anomaly detection system that only uses normal data for the training. When building anomaly detection systems, it is important to capture generalized features based only on the normal data. To solve this problem, first, we propose some algorithms that improve the Deep Autoencoding Gaussian Mixture Model (DAGMM). Second, we propose some algorithms that improves Efficient GAN. Various types of neural networks such as CNN, LSTM, and convolutional LSTM (C-LSTM) are applied to DAGMM, and GMM and C-LSTM are applied to Efficient GAN for effective feature extraction. The experimental results show that each of the proposed methods has effective classification performance for lung sounds, and especially, the combination of convolution and LSTM, and the combination of feature extraction and GMM are effective for any of the models.
Multipoint combinatorial optimization methods based on the hierarchical structure of the solution space are known to fail to achieve an appropriate balance between diversification and intensification due to the degeneracy of the search point set, resulting in degraded long-term search performance. This paper proposes an adaptive combinatorial optimization method that has a mechanism to achieve an appropriate balance between diversification and intensification from the viewpoint of long-term search dynamics while reducing the degeneracy of the search point set. The usefulness of the proposed adaptive combinatorial optimization method is confirmed by numerical experiments using well-known benchmark problems of traveling salesman problems, 0-1 knapsack problems, flow-shop scheduling problems, and quadratic assignment problems.
In order to investigate the potential as a quantitative method of odor evaluation, nonlinear analysis of electrogastrograms was performed. The determinism in the mathematical model generating electrogastrograms (EGGs) during olfactory stimulation was evaluated by the Wayland algorithm. The results showed that the translation error increased with the concentration of the odor samples. The minimum embedding dimension of the attractor reconstructed from the EGGs was estimated by the false nearest neighbors method and was estimated to be 5 to 8 dimensions. However, it was considered necessary to evaluate using the method eliminated the arbitrariness of the threshold value because the minimum embedding dimension varied depending on the threshold value.