◎Bowen Zhang，Masahiro Tanaka（Konan University）

Deep Convolutional Neural Networks(DCNNs) have achieved a state-of-the-art performance in computer vision tasks. But convolution operation can only process a local neighborhood at one time. This work constructs a DCNNs structure to adopt a soft-attention module for capturing the dependencies of spatial position information in each image, which is applied to a Super Resolution task. In addition, we use a few convolution filters with different kernel sizes in parallel for extracting the multi-scale features. In the experimental study, we compare the performance of our model with the traditional Bicubic interpolation method and the deep learning method SRCNN(3 layers DCNNs), where the Peak Signal-to-Noise Ratio index of our model is superior to the other two.

◎Takaya Mori，Kimihiro Nakama，Seiichi Koakutsu（Chiba University）

In recent years, plant factory technology has attracted attention as a technology for stably producing plants. In many cases, leafy vegetables are grown in artificial light-type plant factories. In particular, lettuce has become a major crop. However, lettuce and other leafy vegetables suffer from a physiological disorder called tipburn. Therefore, Convolutional Neural Network is used to automatically identify the tipburn. In this study, we aim to implement the tipburn detection of lettuce produced in an artificial light plant factory on a small Raspberry Pi. However, images for training have different F-numbers than small cameras. As a result of experiments, it was confirmed that the number of images is more important for image classification than the difference in F-number of images.

◎Yasuhiro Makino，Shota Oguma（Hiroshima University），Shuichi Ohno（Osaka City University），Kazuhiro Iwasaki（KOBELCO Construction Machinery Co.,Ltd）

In the construction field, it is important to automatically classify motions of a construction machine to improve productivity and safety. In recent years, the recognition technology using deep learning has been developed for image recognition. However, since a construction machine moves around in a construction field, motion classification by image recognition requires a large number of cameras. Motion classification using sensor data of a construction machine is effective. In this study, we classify the digging motion of a hydraulic excavator by Long Short-Term Memory (LSTM) and reports classification results.

◎Yuuki Toita，Shin Wakitani，Toru Yamamoto（Hiroshima University）

PID control is still widely used in real process systems. Since PID gains strongly affect the control performance, many methods for adjusting PID gain have been proposed. Most of them for systems with single-input and single-output. However, many process systems are multi-input multi-output (MIMO) systems, which have mutual interference. For such systems, a data-driven approach that does not require a model of the system is useful. This study proposes a data-driven multivariable self-tuning PID control system design method. The proposed method employs matrix-style PID gains to control the system considering the effects of mutual interference. The effectiveness of the proposed method is verified by numerical simulations.

◎Masayasu Masuda，Shin Wakitani，Toru Yamamoto（Hiroshima University）

Since most industrial systems have nonlinearities, it is difficult to obtain sufficient control performance by a PID controller with fixed gains. A database-driven PID (DD-PID) control has been proposed as one of effective controller design methods for such systems. The DD-PID controller adjusts PID gains adaptively according to the changes of equilibrium point of a system. Since a lot of computation time are required if the size of database is large, a method of dividing the database has been proposed. However, in this scheme, the number of partitions of a database must be determined by trial and error. This study proposes a method that automatically determines the number of partitions of the database based on the x-means method that uses Bayesian information criterion.

○Shouta Oka，Ribun Onodera，Seiji Saito（Polytechnic University）

Shape memory alloy (SMA) is expected to be used as positioning control actuators and sensors. On the other hand, SMA has non-linear characteristics, and they may affect tracking performance to target position, positioning accuracy, and system stability when SMA is used as the actuator for a positioning control system. Several mathematical models of SMA have been proposed for use in simulation for control systems using SMA actuators and compensator design for non-linear characteristics. However, few consider all of the non-linear characteristics of SMA.
The purpose of this study is to describe SMA actuator model considering minor loops and load variation and to propose a control system using it.

◎Takaya Oguma，Daiki Shimokura，Kenji Kurokawa（Kitami Institute of Technology）

We demonstrated that fiber fuse initiation was suppressed by amplitude modulation (AM) of the input light in a WDM system in both dispersion-shifted fiber and SMF. When the input power was 6 W in SMF, suppression of fuse initiation occurred under AM light with an extinction ratio of 2.0.

◎Tomoharu Konishi，Koichi Iiyama（Kanazawa University）

High-speed and high-accuracy three-dimensional object profiling is realized by using the frequency modulated continuous wave optical ranging system and a two-dimensional Galvano scanner. The Galvano scanner for laser beam scan is continuously scanned to avoid the response delay of the Galvano scanner. In addition, blurring of the three-dimensional object profiling result caused by the change of the illuminating position while data acquisition, is canceled by changing the symmetricity of the triangular modulation waveform of the laser injection current. As a result, the profiling time is 43 times faster than the conventional system using a step-scanned Galvano scanner. 

◎Kano Kajie，Takashi Shibata（University of Toyama），Kenta Shimba（University of Tokyo），Yoshitaka Miyamoto（National Center for Child Health and Development），Tohru Yagi（Tokyo Institute of Technology）

In the ultrasound-based drug delivery system (DDS), the target cells are irradiated with ultrasonic waves to temporarily drill the cell membrane. At the same time, the drug is released from the liposomes and taken up by the cells. However, the optimal irradiation conditions for the ultrasound to maximize drug uptake are often still unclear. In this experiment, we irradiate artificial cell membranes with ultrasonic waves in a solution containing a fluorescent substance that is considered to be a drug. As a result, we have found the irradiation condition that increases the uptake the most. In the future, we will continue to conduct experiments using various irradiation conditions to search for the optimal conditions that maximize the effect of ultrasound irradiation and to apply it to DDS.

◎Kohtaro Hongo，Peng Zugui（Tokyo Institute of Technology），Kenta Shimba（The University of Tokyo），Yoshitaka Miyamoto（National Center for Child Health and Development），Tohru Yagi（Tokyo Institute of Technology）

The insertion of an ion-selective membrane protein into lipid bilayers between droplets of different concentrations generates an electric current. This is expected to be used as a power source for medical devices embedded in the body. However, ion-selective membrane proteins are difficult to synthesize, and a method to control this current has not been developed. Therefore, we have attempted to control the current using carbon nanotubes (CNTs), which are not only ion-selective but also easy to synthesize and chemically stable. In this study, as a preliminary step, we have measured the capacitance of lipid bilayers and confirmed the usefulness of the experimental system. We have also investigated the conductance of CNTs, which suggested that CNTs were inserted into lipid bilayers.

◎Shoichiro Kanno，Zugui Peng（Tokyo Institute of Technology），Kenta Shimba（The University of Tokyo），Yoshitaka Miyamoto（National Center for Child Health and Development），Tohru Yagi（Tokyo Institute of Technology）

Single-walled carbon nanotubes (SWNTs) are hollow, cylindrical carbon materials with high permeability to ions and water molecules. Therefore, SWNTs are expected to be applied by incorporating them into spherical artificial lipid membranes (liposomes). Currently, the mass transport process through SWNTs is evaluated by the change in fluorescence intensity of fluorescent probes in liposomes. However, it is difficult to determine the cause of the change in fluorescence intensity due to their small size. In this study, we propose a new approach using liposomes with a diameter of over 10 µm (giant liposomes). By increasing the size of the liposomes, we can measure the ion penetration and transport processes of SWNTs while easily observing their shapes with a microscope.

○PENGCHENG LI，RONG YANG（Tokyo Institute of Technology），Hironori Nakatani（Tokai University），Tohru Yagi（Tokyo Institute of Technology）

The current study is designed to investigate whether the entrained oscillations could affect performance and information processing in short-term working memory and an arithmetic task. According to some previous studies, it is known  that the flicker frequencies close to 10.2 Hz increases senior people’s memorization of the words which subjects learned before. The current study is not just studying the effects of entrained oscillation of alpha peak waves, but also focuses a broader range at 7.5Hz, 10.5Hz, 16.5Hz. In the experiment, besides a memory task, subjects also need to do the arithmetic task. Current results indicate that the reaction time of the experimental group is faster than a control group. Entrained oscillations may help people become more sensitive and cognitive.

○Mio Yokoyama（Tokyo Institute of Technology），Hironori Nakatani（Tokai University），Tohru Yagi（Tokyo Institute of Technology）

The diagnosis of depression is subjectively made based on the experience and knowledge of psychologists and doctors. An objective evaluation system using physiological indices caused by depression has not been developed yet. Previous studies have suggested that depression is caused by brain structural and functional changes. There have been some efforts to diagnose depression objectively using fMRI and NIRS, but all of them have been limited to adjuncts. Therefore, we aim to develop a new diagnostic system of depression by comparing the EEG of people with high or low depression states. Since it may be difficult to discriminate depression only by EEG in the normal resting state, we will also observe the correlation with the response to visual stimuli.

◎Panida Povarasoontorn（Tokyo Institute of Technology），Takashi Shibata（University of Toyama），Tohru Yagi（Tokyo Institute of Technology）

Ultrasonic hearing is the phenomenon which allows humans to perceive high-frequency sounds by directly stimulating the inner ear via the bone conduction. By modulating the speech with the ultrasound, the speech component can be perceived through the ultrasonic hearing, with the high-pitch ultrasound as a background noise. The bone-conducted ultrasound (BCU) has the potential to be commercially and medically used as hearing devices. However, the characteristics of the modulated BCU have not been extensively studied. Whether or not the BCU hearing device has advantages over the conventional bone-conducted hearing device is still unexplored. Therefore, this research aims to evaluate the characteristics regarding the sound quality and the sound leakage of the modulated BCU.

○Hiroki Noyama，Yuto Yoshikai，Ryuki Iguchi，Tianyi Zheng，Kiyoshi Kotani，Yasuhiko Jimbo（The University of Tokyo）

Cell assembly fires synchronously throughout the brain, and the spike-timings and patterns of cell assembly are thought to be influenced by LFP generated by the interaction of individual neurons. However, the mechanism on how those two are coordinated has not been clarified. In this study, we embed spatial patterns into the fully coupled model and consider a model that oscillates with different spatial patterns when different inputs are given from outside the system. By considering these spatial patterns as the activity patterns of the cell assembly, we can quantitatively evaluate how different input affects the spike-timings and patterns of the cell assembly. we aim to construct a fully coupled neural network model that controls the specific spike-timings and patterns.

◎Yuji Kirihara（Tokyo Institute of Technology），Kenta Shimba（The University of Tokyo），Yoshitaka Miyamoto（National Center for Child Health and Development），Tohru Yagi（Tokyo Institute of Technology）

Engineering skeletal muscle tissue has gained attention greatly. However, the conventional tissues have two major problems; 1) the low density of cells, and 2) the lack of alignment of the muscle fibers. Therefore, we propose a method to increase the density and align the fiber using magnetic nanoparticles and a network-shaped tissue. In this study, embedding magnetic nanoparticles on mouse C2C12 myoblast cells, we evaluate the cellular uptake of the particles and the effects on cells. The current results show that the muscle fiber thickness tends to slightly decrease in C2C12 with internalized magnetite. Therefore, we build a network-like structure with a range of density that does not affect the cell to measure the contraction force.

◎Fumika Moriya，Kenta Shimba，Kiyoshi Kotani，Yasuhiko Jimbo（The University of Tokyo）

In the hippocampus, newborn neurons can be produced even after prenatal development. However, the correlation between neurogenesis and learning and memory function is still unclear. Here, we investigated the relationship between neurogenesis and the dynamics of hippocampal neuronal networks. We administered DAPT to a hippocampal culture with different time courses to control the proliferation and differentiation of neural stem cells. By comparing the spontaneous activity among 3 conditions using a high-density microelectrode array, we assessed the effect of neural stem cells and their development on the dynamics of the hippocampal network over culturing time. Our results give new insight into the role of neural stem cells in hippocampal electrical activity.