Image Processing Laboratory

Ryuichi Oka Professor	Keitaro Naruse Assistant Professor	Yu Nakajima ResearchAssociate

First and foremost, the Image Processing Laboratory engages in research and development of image-based pattern recognition including some areas of Artificial Intelligence, database organization, retrieval, and robotics. More specifically, as can be seen from the background information and the recent research publications of the members of the laboratory, object tracking in a time-varying image, image retrieval from video data, character recognition of telop in a video image and extraction of the Internet community are our current focus. Related to the recent progress of the web, a huge amount of multimedia data without index becomes available to store in our PC's. However, no sophisticated methodology to manage such data has been developed so far. So that we are solicited to attach an index to each data. Our research aim is to develop algorithms to realize automatic annotation to real word data for integrated retrieval of multimedia information. The algorithms include self-organization and transformation among representation of multimedia and feature extraction and recognition of real data. Real world data includes video, still image, speech, music, sound, and text each of which has not been indexed by labels. An example of multimedia integration retrieval called "CrossMediator" developed by ten-year project (1992-2002) RWC of METI (Japan). Some parts of Cross Mediator have been in the commercial market by through a private company. Prof. R.Oka was a chief of group which developed "CrossMediator". Our laboratory will pursue to develop more sophisticated functions which might reveal a new generation of the Internet. Speech retrieval system used in CrossMediator is used for developing new types of commercial products. One of the recent progress is developing a set of algorithms for spotting recognition and tracking objects in a time-varying image. These algorithms enable to realize automatic annotation of video image capturing moving objects. Another recent progress is developing a spotting algorithm for song wave retrieval. A large amount of song wave data is available for anyone along with the progress of storage hardware of music data. But the progress of technology for accessing the song wave data seems quite slow so that we must use conventional and not sophisticated tools. Ournew software can provide users a convenient way to find and retrieve the song wave data by singing an arbitrary part of it which he/she wants to hear. The Internet is regarded as a network composed of virtual communities. Visualization of the community is becoming an important research target. Our lab is developing a sophisticated algorithm based on so-called Associated hyper-linked word space (A-space). We use the algorithm to detect clusters each of which corresponds a virtual community. Visualization of each virtual community is realized by showing the content extracted from the web pages included the community.

Refereed Journal Papers

[naruse-01:2004]	Satoshi Kawai, Keitaro Naruse, Hiroshi Yokoi, and Yukinori Kakazu. An Analysis of Human Motion for Control of a Wearable Power Assist System. Journal of Robotics and Mechatronics, 16(3):237-244, 2004.
	The objective of this study is to build a control system for a wearable power assisting device, which helps people with lifting heavy things, like a person who cannot move alone in daily life. When we lift something heavy, we use both hands. Therefore, a control system that can automatically switch on/off without using hands is necessary. In this paper, we analyze human lifting up motion, and reveal characteristics of the motion, to find the suitable timing for switching the wearable power assisting device. Results show that the automatic switching can be realized.
[naruse-02:2004]	Ikuo Suzuki, Masaru Fujii, Keitaro Naruse, Hiroshi Yokoi, and Yukinori Kakazu. Acquisition of Adaptive Behavior for SMA-Net Robot using Chaotic Neural Networks. Journal of Robotics and Mechatronics, 16(4):411-419, 2004.
	In this paper, we propose control for the SMA-NetRobot, a flexible structure consisting of many units and shape memory alloy (SMA) springs. To generate greater force for movement, more than oneSMA spring is required. Since SMA springs are driven by thermal transition, controlling individual spring heating patterns is important in SMA-Net Robot behavior. It is a problem in controlling SMA spring that its detailed control is difficult because of the nonlinearity. We propose methodology that arranges heating and cooling as a rhythm pattern memorized by many chaotic neural networks (CNNs). To renew connecting weights in the network, we use the modified dynamic learning method (DLM) in online learning. The results of computational experiments showed that the SMA-Net Robot with the proposed control generates movement automatically.
[naruse-03:2004]	Keitaro Naruse, Satoshi Kawai, Hiroshi Yokoi, and Yukinori Kakazu. Design of Wearable Power Assist Device for Lower Back Support. Journal of Robotics and Mechatronics, 16(5):489-496, 2004.
	In this paper, we propose a mechanical device for supporting the lower back, which is often injured due to carrying heavy loads because of the large compression applied to the lower back that may injure spinal discs. We develop a power-assist device for supporting one degree of freedom (DOF), which is flexion and extension of the lower back, for reducing compression and assisting lifting. We evaluate the effect of the device in experiments that verify that the proposed power-assist device effectively protects against lumbago.
[naruse-04:2004]	Makoto Oya, Keitaro Naruse, Masahiko Narita, Taku Okuno, Masahiro Kinoshita, and Yukinori Kakazu. Loose Robot Communication over the Internet. Journal of Robotics and Mechatronics, 16(6):610-618, 2004.
	As robot technology is disseminated from special fields such as factory production to everyday fields such as home use, community work, and entertainment, working space of robots also expands from dedicated area such as factory sites to ordinary living space. Combining this technology shift and the popularized public Internet suggests the emergence of new robot applications. The Internet, however, as its nature, is not suitable for tight coupling of processes among robots and computers. An approach different from simple expansion of current robot communication technology is thus necessary. After discussing the essentials of communication, i.e., closed and open networks, and tight and loose couplings of processes, we propose combining Web Services technology and autonomous robot technology. We present several cases to examine this approach. The applicability of Web Services to robot communication is confirmed through implementation. Feasible applications are studied using application instances combining autonomous robots with the Internet. Feasibility is also examined fromWeb Services technology point of view.

Referred Proceeding Papers

[naruse-05:2004]	Satoshi Kawai, Keitaro Naruse, Hiroshi Yokoi, and Yukinori Kakazu. A Study for Control of a Power Assist Device - Development of an EMG based controller considering a human model -. In Proc. of 2004 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS2004), pages 2283-2288. IEEE/RSJ, Sept. 2004.
	Apower assisting device for lower back flexion and extension, when carrying a heavy load is presented in this paper. Whenpeople lift up something heavy, they use both hands. So the development of a controller without using hands is necessary.We attempt to control the device by a voluntary human motion. First, lifting up and putting down motions are analyzed using a human model. Results show that torques in a hip joint and a knee joint are characteristic for control the device. It means that EMG signals in front and back thigh muscles can become feature values because the muscles located there connect the hip joint and the knee joint. After that, a controller of the device using the EMG signals is developed. Finally, artificial neural networks(ANN) are introduced as a solution of the problem of individual differences.
[naruse-06:2004]	Keitaro Naruse, Kazutoshi Uehara, and Yukinori Kakazu. Coupled Evolution of Behavior and Morphology in Virtual Creatures. In Proc. of the 4th International Conference on Advanced Mechatronics (ICAM2004), pages 84-89. JSME, Oct. 2004.
	This paper presents the investigation on the simultaneous evolution of the behavior and morphology. For the purpose, we developed a robot model, which is simple but has high representation ability. We proposed the coupled evolution method for the model. The computer simulations were made, and we confirmed the followings: The coupled evolution can decrease the stagnation of evolution. When the coupled evolution is success, the morphology is changed that the forces generated by muscles are eAEciently converted to the motion. It is appeared in the increase of the speed, which is shown in the higher fitness value by the coupled evolution. We do not see the direct relationship between the model complexity and the success of the coupled evolution.
[naruse-07:2004]	Keitaro Naruse, Tomohiro Maeda, and Yukinori Kakazu. Speech Recognition and Template Learning for Bone Conduction Microphone. In Proc. of the 4th International Conference on Advanced Mechatronics (ICAM2004), pages 181-186. JSME, Oct. 2004.
	The goal of this research is to build the speech recognition method that can bear environmental noise using a bone conduction microphone that has been used for many years to secure voice communication under high noise environment and speech recognition equipment. The bone conduction microphone has a disadvantage of narrow bandwidth, so a conventional speech recognition method for a normal microphone can not be simply applied. In this paper, we propose a speech recognition system reflecting a characteristic of the bone conduction microphone with learning a voice template. Experiment results show that the proposed system works well under the high noise environment.
[naruse-08:2004]	Ikuo Suzuki, Keitaro Naruse, and Yukinori Kakazu. A Study on Adaptive Navigation System using Nonlinear Neural Networks. In Proc. of the 4th International Conference on Advanced Mechatronics (ICAM2004), pages 316-321. JSME, Oct. 2004.
	In this paper, we propose an emergent system in which an autonomous mobile robot can acquire environment oriented behavior. In the general robotic engineering, a system designer gave the model of environment and sensory-motor commands beforehand. However, we think that the autonomous robot has to be developed through the interaction between robot's behavior and the environmental information by itself. So, we construct the controller for the autonomous robot to acquire the adaptive behavior with the Chaotic Neural Networks (CNNs). Furthermore, we make use of the Dynamic Learning Method (DLM) as an on-line learning method. The results of the computational experiments show the chaotic search of this network plays an important role for the acquisition of the adaptive behavior.
[naruse-09:2004]	Satoshi Kawai, Keitaro Naruse, Hiroshi Yokoi, and Yukinori Kakazu. Development of Small Size Controller for Control of a Wearable Power Assist Device. In Proc. of the 4th International Conference on Advanced Mechatronics (ICAM2004), pages 407-412. JSME, Oct. 2004.
	When a person lifts up or puts down heavy loads, large load occurs at the intervertebral disk. Large load on the waist causes lumbago, which makes people hard to work. In this paper, a power assist device which helps people lifting up or putting down heavy loads is developed, and a controller which adapts to human motion automatically is proposed. First, human motions are analyzed and feature values to control this device are investigated. Furthermore, we introduce an embedded CPU to the controller in order to walk around easily even while walking.The experimental result shows our proposed small size controller works well for some subjects.
[naruse-10:2004]	Masayoshi Saito, Keitaro Naruse, and Yukinori Kakazu. Express of Pet Robot's Emotion by Motion of its Head. In Proc. of the 4th International Conference on Advanced Mechatronics (ICAM2004), pages 430-435. JSME, Oct. 2004.
	Autonomous behaviors are necessary for the entertainment robot which gives richness to humanbeings. Especially for the pet robot, autonomousbehaviors as if it has the emotion are necessary. The pet robot expresses one's own emotion in order to realize this, and the emotion must be communicated to the human. In this paper, by analyzing the relation between features of the motion of the pet robot and the impression which the motion gives, it is enabled that the emotion of the pet robot is expressed by the motion of its body. To begin with, the impression which the motion of the pet robot gives by questionnaire method is surveyed. Next, the survey result is analyzed using statistical method. Then, both relation is clarified. By this, it was proven that any emotion could be expressed in the simple motion only of the head. And, it was possible to also express the emotion in the similar motion in the object without the face which formed the head on the specific emotion.
[naruse-11:2004]	Masahiko Narita, Keitaro Naruse, and Makoto Oya. RoboLink: ARobot Collaboration Protocol based on Web Services. In Proc. of the 4th International Conference on Advanced Mechatronics (ICAM2004), pages 442-448. JSME, Oct. 2004.
	Studies and standardizations of remote-robot-control have mainly been focusing on real-time communication and precise control. As the robot technology extends to social activities such as entertainment and home-use, a looser communication method is required to realize robot computer and robot-robot collaboration in the open/public network space, which does not have to be real-time or precise. The RoboLink Protocol was developed based on standardized protocols for such collaboration. The Web services technology was adopted as its base to realize this looseness. This paper outlines the design principles and the resulting specification of the RoboLink Protocol. It also reports experimental results that show feasibility of the RoboLink Protocol and business applicability.
[naruse-12:2004]	Makoto Oya, Keitaro Naruse, Masahiko Narita, Taku Okuno, Kenichi Mori, Masahiro Kinoshita, andYukinoriKakazu. Loose Robot Collaboration in the Public Internet Environment. In Proc. of the 4th International Conference on Advanced Mechatronics (ICAM2004), pages 454-461. JSME, Oct. 2004.
	The robot technology is extending to social activities such as home-use, community work and entertainment. By combining it with the popularized public internet, it is anticipated a new application field will come out. However, the public internet, as its nature, does not permit a tight coupling of processes among robots and computers. A different approach is necessary regarding communication protocols and forms of robot applications. This paper proposes the combination of the Web Services technology and the autonomous robot technology is the proper approach. Several cases are studied to examine the feasibility of this approach. Applicability of the Web Services to robot communication is verified through the implementation practice by the Robo Link protocol. The feasible applications are studied using application instances combining autonomous robots with the internet. Applications derived from the Web Services technology are also studied.
[naruse-13:2004]	Masashi Furukawa, Daisuke Kobayashi, Michiko Watanabe, Masahiro Kinoshita, Takashi Kawakami, Norihiko Takatori, Katsumi Hama, and Keitaro Naruse. Vehicle Path Planning by Use of SOM. In Proc. of the 4th International Conference on Advanced Mechatronics (ICAM2004), pages 132-136. JSME, Oct. 2004.
	The self-organizing map method (SOM) is applied to the vehicle path-planning problem (VP*3). It is known that the SOM is capable of solving VP*3, such as the traveling salesman and path planning problems. However, SOM often makes intersection edges, applying to the one-way path planning and n-vehicle path-planning problems. This study proposes a new method, to which a multi-knot neurons concept is supplied for removing intersection edges, in solving these problems. Somenumerical experiments verify that the proposed method obtains good approximate solutions.
[naruse-14:2004]	Masahiro Kinoshita, Takashi Kawakami, Makoto Oya, Keitaro Naruse, Masashi Furukawa, Michiko Watanabe, and Yukinori Kakazu. The Adaptive System for Collective Behaviors of Multi-Robots. In Proc. of the 4th International Conference on Advanced Mechatronics (ICAM2004), pages 553-558. JSME, Oct. 2004.
	This paper proposes the new methodologies of smart control for collective agents based on the adaptive systems. The smart control means that the collective agents make a decision of their strategies under the dynamic and illknown environments. The strategies are acquired by adaptation through interaction between the collective agents and the environments. This acquisition mechanismsare so-called adaptive strategic mechanism for collective agents in this paper. This type of agent can sometimes be considered as mobile robots in the robotics field. However, the proposed methodologies of the control and emergence of collective patterns for the collective agents are applicable to other engineering fields such as information network, social behavior.
[naruse-15:2004]	Keitaro Naruse, Masahiro Kinoshita, Masashi Furukawa, and Ryuichi Oka. Aggregation and Distribution of Social Agents. In Intelligent Engineering Systems through Artificial Neural Networks Vol. 14, pages 189-194, NewYork, Nov. 2004. ASME Press.
	The objective of this paper is to present acontrol method of self-organized formation such as aggregation and distribution of agents in a non-metric space. In this paper, the self-organization is regarded as finding multiple solutions simultaneously in a multi-modal function in a discrete space. For the search problem, we introduce the concept of the particle swarm optimization (PSO). However, it has been knownPSOconverges toomuch to a single solution. We solve the problem by giving social structure to the agents. We investigate and discuss how the social structure a??ects to the self-organization of the agents as well as the search ability. Computational experiments are conducted and the results show that the proposed agent system can find a good solution while the aggregation and distribution can be controlled by a parameter.
[naruse-16:2004]	Satoshi Kawai, Keitaro Naruse, Hiroshi Yokoi, and Yukinori Kakazu. A Study for Control of a Wearable Power Assist System - Recognition of Human Motions by Surface EMG Signals -. In Intelligent Engineering Systems through Artificial Neural Networks Vol. 14, pages 767-772, New York, Nov. 2004. ASME Press.
	This paper shows the development of a wearable power assisting system, which helps people lifting a heavy object like a person who can not move alone in daily life. Whenwe lift something heavy, we use both hands. That is why a control system that canbe automatically switch on/off without using hands is necessary. In this paper for utilizing the automatic control, we analyze lifting motions and other motions andEMG signals detected from a body surface in order to measure the human intention. For this purpose, by utilizing artificial neural networks (ANN) for control of the wearable power assist device, is proposed, and the detail design of ANN is discussed. The experiment results show the adaptability of the proposed ANN controller, which can adapt to the difference of individuals, and it can recognize the five patterns of human motions in the daily life.
[naruse-17:2004]	Masashi Furukawa, Michiko Watanabe, Masahiro Kinoshita, Katsumi Hama, Takashi Kawakami, Norihiko Takatori, Keitaro Naruse, and Yukinori Kakazu. Collision Avoidance for Multi-AGVby Learning Collision. In Intelligent Engineering Systems through Artificial Neural Networks Vol. 14, pages 143-148, NewYork, Nov. 2004. ASME Press.
	This paper represents anew collision avoidancemethod, which is based on reinforcement learning for multiple AGVs. In order to realize it amongthem, at first, the multiple AGVs learn how to make collision. Q-Learning is employed to them for acquiring their collision course knowledge. Then, the multiple AGVs oppositely make use of the knowledge to avoid collision. The reason why the indirect decision way is adopted is, that pursuing other AGVs is easier than avoiding one by learning. Numerical experiments are attempted in case when two to six AGVs meet at a same point. Results verify that all AGVs can stray out their course to avoid collision among the multiple AGVs and return to their own course.
[naruse-18:2004]	Michiko Watanabe, Masashi Furukawa, Masahiro Kinoshita, Takashi Kawakami, Katsumi Hama, Norihiko Takatori, Keitaro Naruse, and Yukinori Kakazu. Generation of Pareto Optimum Map for Job-shop Scheduling. In Intelligent Engineering Systems through Artificial Neural Networks Vol. 14, pages 917-922, New York, Nov. 2004. ASME Press.
	A solution of Job-Shop scheduling, which is one of typical examples of combinatorial problems, is represented as combination of multiple indices. Therefore, a unique solution is not determined; instead, a set of solutions which satisfy a certain constraint is generated (pareto-optimum solutions). In this paper, to improve manually a solution made by an automatic planner, visualization of those solutions is carried out.
[oka-01:2004]	Yuya Iwasa and Ryuichi Oka. Spotting recognition and tracking of a deformable object in a time-varying image using two-dimensional Continuous Dynamic Programming. In Proc. of the fouth Inter. Conf. on Computer and Information Technology, pages 33-38. IEEE, Sept. 2004.
	An object is represented by a two-dimensional image with concave- and convex shape and called a reference image. The reference image is spotted and recognized simultaneously. The spotted image at time t becomes a reference image at time t+1 and so on. This incremental process is equivalent to the tracking of the object which is allowed to change the shape of it continuously.

Unreferred Papers

[naruse-19:2004]	Keitaro Naruse, Tomohiro Maeda, and Yukinori Kakazu. Design of speech command for bone conducted microphone. In 2004 JSME Conference on Robotics and Mechatronics, Jun. 2004.
[naruse-20:2004]	Satoshi Kawai, Keitaro Naruse, Hiroshi Yokoi, and Yukinori Kakazu. A study for Control of a Wearable Power Assist System - Recognition of Human Motions Utilizing Arti??cial Neural Networks-. In 2004 JSME Conference on Robotics and Mechatronics, Jun. 2004.
[naruse-21:2004]	Keitaro Naruse, NorihikoTakatori, Masahiro Kinoshita, Masashi Furukawa, and Ryuichi Oka. Effect of Connectivity in Particle Swarm Optimization. In 2004 Autumn JSPE Conference, Sept. 2004.
[naruse-22:2004]	Keitaro Naruse. Agent Connectivity and Formed Artificial Society. In 2004 SICE System Integration Division Annual Conference, Dec. 2004.
[naruse-23:2004]	Keitaro Naruse, Norihiko Takatori, Masahiro Kinoshita, and Masashi Furukawa. Search in network space by social topology agents. In 2005 Spring JSPE Conference, Mar. 2005.
[oka-02:2004]	Yuya Iwasa and Ryuichi Oka. Spotting recognition and tracking of a deformable object in a time-varying image using two-dimensional Continuous Dynamic Programming. In Technical Report of IE2004-23,PRMU2004-47, MVE2004-24, pages 19-24. IEICE, July 2004.
	When the background is varying or a tracking target is deforming, a unified processing composed of segmentation, recognition and tracking is difficult to realize by conventional methods. An iterative method is proposed by a combination of spotting recognition with a reference image for tracking a target and making as egmented image as a reference image for the next frame. Simultanous processing of recognition and segmentation is called spotting recognition. 2DCDP performs spotting recognition of image. An extended 2DCDP is used to deal with multiple objects of image with arbirary shape. Experimental results showed that the method worked well.
[oka-03:2004]	Yuuichi Yaguchi and Ryuichi OKA. Song wave retrieval based on frame-wise phoneme recognition. In Technical Report of ISP2004-50, pages 19-24. IEICE, August 2004.
	A song wave retrieva method is proposed. Both song wave data and a query song wave data are transformed into phoneme sequences by frame-wise labeling of each frame feature. Applying a spotting algorithm called Continuous Dynamic Programming to these phoneme sequences, we can detect a set of similar parts in the song database each of which is similar to a query song wave. Song retrieval rate hits 82% though the rate of phoneme recognition is relatively low 25%.

Grants

[naruse-24:2004]

Keitaro Naruse. Grants-in-Aid for Scientific Research, 2004-2007.

Academic Activities

[naruse-25:2004]	Keitaro Naruse, 2004. Program Committee, The Fourth International Conference on Advanced Mechatronics (ICAM2004)
[naruse-26:2004]	Keitaro Naruse, 2004. Local Arrangement Committee, The Fourth International Conference on Advanced Mechatronics (ICAM2004)
[oka-04:2004]	Ryuichi Oka, April 2004. Member of Organizing Committee of QCAV2005.
[oka-05:2004]	Ryuichi Oka, May 2004. Reviewer of submitted papers, the Institute of Electronics, Information and Communication Engineers
[oka-06:2004]	Ryuichi Oka, April 2004. Member of editorial committee, Japanese Society for Artificial Intelligence
[oka-07:2004]	Ryuichi Oka, Oct. 2004. Member of International Advisory Committee and Reviewer of submitted papers, 4th World Congress on Industrial Process Tomography
[oka-08:2004]	Ryuichi Oka, April 2004. Reviwer of submitted papers, the fouth Inter. Conf. on Computer and Information Technology

2006 THE UNIVERSITY OF AIZU ALL RIGHTS RESERVED