- Life Sciences
- Information and Communication
- Nanotechnology / Materials
- Manufacturing Technology
- Human and Social Sciences
- Tourism / Community development
- Arctic Research
- Social Infrastructure
- Open Facilities
Information and Communication: 37
Nano Knowledge Exploration Project
Knowledge discovery from experimental records of nanocrystalline devices
In this research, we are studying knowledge management to extract and organize useful information for device development from experimental records and papers compiled in the process of research and development of nanocrystal devices.
In this presentation, we propose an experiment record management system based on interviews with researchers in actual nanocrystal device development. The system proposes an integrated method for the management of records of parameters used in experiments that have been stored and recorded separately, as well as the resulting experimental records. In addition, by extracting information from the paper summarizing the final experiment, we propose a method for analyzing the purpose and characteristics of a series of experiments conducted by researchers in detail, and for using it as the basis for discussing similarities between various cases. With this method, we are proposing a method of extracting useful information from unknown papers by using machine learning methods on a corpus of information manually extracted by a few people.
A method to evaluate radio wave propagation characteristics of in-vehicle wireless access services by large-scale electromagnetic field analysis
Toward optimal design of wireless environments
We have conducted research in various fields where radio waves are applied, including the evaluation of complex and special propagation environments in airplanes and passenger railroad cars, internal exposure of the human body to radio waves, electromagnetic interference evaluation and mechanism estimation regarding body medical devices implanted in the body, and evaluation of electromagnetic field leakage in wireless power supply devices for electric vehicles.
The radio propagation environment inside a vehicle is a special environment that differs from the conventional propagation model due to multiple reflections caused by the surrounding metal and the presence of fixtures and passengers inside. It is therefore necessary to evaluate the characteristics of radio wave propagation, including the effects of absorption and scattering of radio waves by the bodies of passengers, to estimate the quality of the wireless connection under actual operational conditions. However, it is difficult to evaluate the propagation characteristics by actual measurement or simple numerical analysis (e.g., ray tracing). This study addresses modeling of the propagation environment in vehicles, which has been difficult in the past, and realizes a simulation method in a very large analysis space by using a supercomputer.
Promoting research and development through academic cloud collaboration
We promote research to create an academic intercloud that links nationwide cloud systems, and conduct joint research on cloud-related technologies such as the optimization of resource allocation in intercloud environments and linkage between supercomputers and interclouds.
Hokkaido University Information Initiative Center has established the Hokkaido University Academic Cloud, one of the largest academic clouds in Japan, and provides cloud services to researchers nationwide, including those concerning virtual and physical machines and their cluster systems, high-speed high-capacity cloud storage services, and machine learning and big data processing systems for research and development. We are also promoting research on fundamental technologies to realize nationwide cloud system collaboration and system construction to support researchers. Specific examples include the development of infrastructure technologies for cloud collaboration (authentication collaboration, etc.) and the construction of a test system (Fig. 1), resource allocation optimization in an intercloud environment, and the realization of a large-scale design optimization framework on a nationwide scale by linking supercomputers and intercloud infrastructure (Fig. 2). We are also conducting joint research with universities, research institutes and companies across the country.
Accelerometer to Identify Where You Stumble
To prevent falls in elderly people
To avoid serious injuries due to falls, we studied a system that can search places with a potential danger of stumbling, which is a sign of a fall, based on people’s daily activities. The system uses an accelerometer embedded in sandals to identify the stumbling point, and an infrared sensor network on the ceiling to identify the location of the stumbling point.
According to a survey by the Tokyo Fire Department in 2014, approximately 80% of all cases of emergency transportation of elderly people are due to accidents involving falls. Since their consciousness cannot keep up with their declining physical abilities, they stumble over small steps, footwear, or clothing. Much research is done on the detection of falls, but it is too late after the actual fall. Therefore, we decided to eliminate the causes of falls by detecting the areas where people often stumble. Wearable devices have problems with forgetfulness and psychological resistance to wearing it, while non-wearable devices such as surveillance cameras have problems with blind spots and privacy protection. In this study, an acceleration sensor is attached to normal sandals to detect a stumble, while a network of infrared sensors installed on the ceiling works in conjunction to identify the location of the stumble. In our experiments, we were able to distinguish falls easily, but the accuracy rate of distinguishing stumbling from normal walking is currently only about 1/4, so we hope to improve the accuracy in the future.
Advanced Optical Communication Technology, the Core of the Next-generation Ultrahigh-speed Communication Network
Aiming at spectacular advancement of information and communication networks
To realize spectacular advancement of information and communication networks of which the capacity has to increase by a factor of 1,000 in the next 20 years, we are working to develop photonic infrastructure technology that will be the core of next-generation ultrahigh-speed communication networks.
We are studying new structured optical fibers to overcome the limitation of existing optical fibers, optical fiber application technologies to ensure safety and security, ultra-compact optical circuits to support opticalization, and optical simulators to support the design of optical fibers and optical circuits.
An Idea-supporting Multimedia Search System
An information retrieval system that organically links images, video and other data to help searchers find inspiration and ideas.
The idea-supporting multimedia search system organically links unstructured data such as images, music and video, extracts inherent similarities and effectively presents them to searchers to help them find ideas and inspiration.
We have succeeded in establishing associations and similarities between different media, and developed an associative search scheme that takes ambiguity of multimedia information into consideration (fused search). We have also realized a new search engine and interface by quickly introducing modeling of personal preferences through user networks and visualization of similarities in preferences through user interfaces (personal adaptive search). Use of the search engine and interface enables a completely new search that effectively utilizes the polysemy and ambiguity inherent in multimedia contents.
AR Communication System
Avatar-based augmented reality group communication through sharing of terminal position and posture information
By using avatars as intermediaries, it is possible to communicate beyond the constraints of time and space. In this system, AR communication is realized by sharing the position and posture information of participants in group communication and reflecting it in the behavior of avatars visible from each participant's terminal.
Since conventional avatar-based communication systems are based on one-to-one communication, it is difficult to recognize the position and posture information of each physical person participating in a communication field of three or more people where virtual and real worlds are mixed, and to control the behavior of individual avatars accordingly, while maintaining the integrity of the entire field.
In this research, we have developed an AR communication system that shares the position and posture information of physical people (e.g., smartphone terminals) participating in a communication field through recognition of a common AR target and network linkage between terminals, and reflects this information in the behavior of avatars visible from each terminal. The avatars know who and where the participants are in the communication field, and the avatar changing its pose following the movements on one participant's terminal can be seen by other participants from their own viewpoints.
Boundary Element Analysis Framework and Distributed H-matrix Method
Realizing cutting-edge, large-scale, high-precision analysis
We have developed a software framework for high-performance boundary element analysis in a parallel computing environment. We have also developed a distributed parallel H-matrix library, which is effective to accelerate a BEM analysis, and evaluated its performance on programs in various applications.
In this study, we developed a parallel boundary element analysis framework as part of the JST CREST project on “Application Development and Execution Environment with Automatic Tuning Mechanism.” By using this framework, it is possible to develop boundary element analysis programs for large-scale parallel computing systems with a small programming cost. We also developed a library of H-matrices, which approximates dense matrices with less data using low-rank approximation. This library can be used for analysis of N-body problems in addition to boundary element analysis, and is already being used for earthquake simulation on supercomputers. There is no other H-matrix library that supports hybrid parallel processing utilizing multiple threads and processes, and this is a unique feature of this research.Takeshi Iwashita Professor
Communication Robot System
Social space recognition system using dialogue activity and attention guidance system and multiple robots
By calculating the level of activity of a dialogue between people, a robot can recognize the intensity of that dialogue space and adapt its behavior to the context. Furthermore, by applying this mechanism to the behavior of multiple robots, it will become possible to guide the user's attention.
Our dialogue activity calculation system calculates the real-time activity level using information such as the distance between the interactants, voice data, and body movements. By using this activity level, the robot can determine whether it is allowed to enter the dialogue space or interrupt the dialogue, and can take contextually adaptive actions. Furthermore, by having multiple robots act in a way that increases the dialogue activity level for each other's actions, the user's attention (e.g., gaze) can easily be guided. Such a system for generating robot behavior using the level of dialogue activity has not been included in conventional research on social robots, and can be applied to robots at reception desks and home robots for households.
Communication-avoidant Matrix Calculation Algorithm
Research and development of algorithms suitable for massively parallel computers
With the spread of massively parallel computers, it has become increasingly important to reduce the communication time associated with parallel computation. In this research, we aim to improve the performance of matrix computation algorithms by using an approach called “Communication Avoiding (CAA).”
With parallel processing using large-scale parallel computers, the data communication time is often more important than the computation time. In particular, the large communication latency (the cost incurred regardless of the amount of data to be communicated) has become a problem, and there is a strong need to reduce the communication frequency (communication avoidance). We are reviewing existing matrix computation algorithms from the viewpoint of communication avoidance, and are researching and developing new algorithms for massively parallel computers that reduce the communication frequency.
Compact Aerial Video Display System
Direct Touch Visual Presentation System with which images look as if they are floating on the desktop and can be touched directly
We have developed a small video display system that displays images such as 3DCG, which are conventionally shown on a display, in the air. Users can directly reach out and touch the images that seem to float in front of them on their desktop.
Improvement in the performance of stereoscopic displays and head-mounted displays (HMDs) has led to improvement in the quality of 3DCG and virtual reality (VR) experiences. With this research, we are studying the next generation of information interfaces. Aerial video display technology does not require displays, goggles, or other equipment in the space in front of the user. With the technology, 3DCG can be displayed in real space, and the user does not have to enter virtual space, thus it is possible to present information naturally, both in terms of equipment and visuals. Furthermore, the user can directly reach out to where the image is displayed, and can touch and move the image. In addition to realizing this aerial video display system in a desktop size, we are also researching the technology to create video content suitable for it.
Developing Interfaces that Read Users’ Intentions
Controlling robots and mice freely
We are developing interfaces that make it possible for users and machines to learn from each other and to operate machines as the users intend. It also facilitates the control of robots and other machines, and input with pointing devices such as mice and trackballs.
For a user to operate a multi-degree-of-freedom robot such as a humanoid robot, it is necessary to remember which commands correspond to which operations, and the more commands there are, the harder it becomes for the user. Since it is usually difficult to prepare a set of commands that are easy to remember and use for all, there is no guarantee that the interface created will be easy to use. With this research, we read the user's intentions through the interaction between the user and the machine, and construct an interface that the user can operate intuitively. As a result, we can develop easy-to-use interfaces that match the characteristics of individual users. We are also applying this technology to the development of an air mouse and air trackball that can be operated without a device by sensing the movement of the hand operating the mouse or trackball.
Development of a High-speed Japanese Input System for Mobile Devices
Development of a fast and easy Japanese input method for smartphones and other mobile phones
There is still no fast and easy way to input Japanese on mobile devices (smartphones, etc.). With this research, we are developing a system that halves the number of keystrokes normally required to input Japanese characters by using a powerful learning function that matches the respective lines in the Japanese syllabary with numbers.
This system translates strings of numbers input by degenerated keywords into a Japanese sentence with mixed Chinese characters and Japanese phonetic characters. Through degenerated keyword input, one line in the Japanese syllabary is matched with one number to allow fast and easy input. This system provides words by using inductive learning to compare the strings of numbers and the proofread translation results. Therefore, even if the dictionary is empty, the system can generate a dictionary adapted to the target field. However, input of a sequence of numbers from degenerated vowel information may cause ambiguity. To solve this problem, this system uses information on adjacent characters, top-level words and position presumption information. The use of top-level words and position presumption processing increases the number of words acquired, and the adjacent character information enables the conversion that takes connections between words into account.
Development of an Electronic Holographic HMD Device
Ideal 3D image display
We are developing a head-mounted 3D display (HMD) device using electronic holography. This is a compact, lightweight and practical device realized by using a new computational algorithm and a newly designed optical system.
Electronic holography-based display devices can display stress-free 3D images that are compatible with human visual physiology. Using computational correction, the head-mounted display (HMD) developed with this research replaces the complex optical system that had been required in the past, with a simple optical system. As a result, we have achieved the smallest and lightest electronic holography system in the world at low cost. As shown in the figure below, the hologram image can also be displayed to fit in the depth.
Unlike conventional HMDs, this HMD does not cause users stress due to a shift in depth focus.
Development of Mathematical Algorithms for Biomedical Optical Imaging
Development of a mathematical model for light propagation model inside biological tissues
A highly accurate and computationally efficient light propagation model is necessary for the progress of biomedical optical imaging. In this study, we have succeeded in constructing a fast solution method for the radiative transfer equation that describes light propagation with high accuracy. We are working on the advancement of an optical diagnosis and treatment using the proposed method.
In this study, we are constructing a mathematical algorithm for biomedical optical imaging based on the radiative transfer equation. Our goal is to develop an imaging technique with excellent image resolution that can be applied to biological tissues and body parts where conventional imaging based on mathematical models cannot be applied. Until now, the numerical computational burden of the radiative transfer equation has been enormous, limiting its applicability to small-sized organisms. In this study, we have succeeded in developing a highly accurate and computationally efficient light propagation model by coupling the radiative transfer equation and the photon diffusion equation. Optical imaging based on the developed light propagation model can be applied to various biological tissues and sites. Currently, we are working on applying the model to the optical diagnosis of thyroid tumors in the human neck and the in-vivo evaluation of optical property values in biological tissues.
Estimating the State of Radio Waves Using the Compressed Sensing Method
Toward highly accurate location estimation and channel prediction
The compressed sensing method is a method to find a solution under certain conditions from a smaller number of observation data than the number of unknowns required. In this study, we use compressed sensing for estimating the direction of arrival of radio waves, to predict the channel, and detect scatterers.
It is usually impossible to specify unknowns if their number among observation data is smaller than the number of unknowns that need to be found. However, in case the majority of unknowns are zero, it is sometimes possible to obtain the exact solution. Compressed sensing is a method for obtaining an accurate solution while minimizing the number of observations by using this property. In our laboratory, we are investigating the application of this method to high-precision estimation of the direction of arrival of radio waves as shown in Fig. 1, a method of channel prediction by dividing the incoming wave into elementary waves using this method (Fig. 2), and scatterer detection using the compressed sensing used in radar systems (Fig. 3).
Event Information Recommendation System
A system that collects data from a few weeks before an event to the day of the event and recommends appropriate event information.
Although event information was only valid for a short time and it was hard to handle it with conventional information recommendation technology, we have developed a flexible recommendation method by combining multiple factors such as user interest and geographic characteristics.
The system estimates the genres and information sources that a user prefers based on the user’s past information browsing history, and assesses the event information that the target user is interested in by referring to the browsing trends of users with similar interests. It also takes into account the geographic characteristics of the user and finally presents the information to him/her. The timing of information distribution is adjusted throughout the system so that the overall system performance can be improved.
Gesture Recognition System Using Ultrasound
A system that can recognize gestures, including the internal state of the human body, based on the response to ultrasonic waves propagated in the human body.
We are developing a method of recognizing gestures, including the internal condition of the human body, such as the amount of force applied to an object, by transmitting ultrasonic waves to the human body using microphones and speakers in smartwatches or other devices and analyzing the responses.
Our gesture recognition system uses a method called active acoustic sensing, which sends out sound waves and analyzes the state of the object using the returned response. This method enables gesture recognition including the internal condition of the human body, such as the amount of force applied, which has been difficult to recognize with conventional accelerometers. It is intended to use this method in a wearable computing environment, and it is unique in that it can simultaneously acquire both externally visible gestures and internal gestures using microphones and speakers that are installed in many devices. For example, we envision that our method can be applied to gesture operations when both hands are occupied in a wearable environment, and to the lifelog of the amount of force applied under stress or other situations.
High Value-added Media Information and Communication Technology
High value-added media information and communication technology using sub-channel data communication with information hiding technology
We are researching high value-added media information and communication technology that can add new functions while maintaining compatibility with standard formats, by using information hiding technology that is usually used as an information security method.
This research is aimed to add value to media information and communication technology using sub-channel data communication with information hiding technology.
As an application example, this study investigates high-fidelity sound of the voice on the telephone. With the proposed method, information for widening the bandwidth is embedded in the voice data in advance at the transmitter side, so that the voice can have high-fidelity on the receiver side.
This research also investigates a method for detecting tampering by embedding information that can guarantee the originality by image data. Figures 1 to 3 show an example of the proposed method for detecting tampering with a car license plate. When an image is tampered with, the corresponding part of the image is revealed, and the proposed method will not only detect the tampering, but also the tampered part.
High-precision Acoustic Position Recognition, Time Synchronization, Selective Flickerless Visible Light Communication
Submillimeter-order position measurement and its deployment
By integrating ranging technology that is more accurate than conventional methods by double digits and original time-synchronization technology using illumination, we can quickly and accurately estimate the 3D position and velocity of mobile terminals and robots. The system also realizes selective flickerless visible light communication and position-dependent information distribution to specific moving objects.
To accurately obtain the position of a user in a room or a moving object in real time, we have proposed a high-precision time reference point setting method called the phase-matching method (ranging error of 0.03 mm). Based on this technology, we have developed a smartphone users’ gesture recognition system and a robot tracking system. We have also achieved microsecond-order time synchronization using a camera-equipped mobile terminal and an original algorithm. The integration of LED modulation and terminal location information will allow us to distribute location-dependent information and conduct flickerless visible light communication using indoor lighting.