Regardless of complex, unfamiliar, and dynamically-changing environments, living creatures, especially humans, can adapt to them in real-time and behave appropriately. To understand the adaptation mechanisms, and to realize intelligent artificial systems based on the mechanisms, we are (1) investigating human visuomotor learning process, (2) developing a computational model of neural networks with high-adaptive, and (3) designing adaptive personal robots and human interfaces.


Reinforcement Learning models for understanding human motor learning and cognitive learning processes

Imitation learning and reinforcement learning works as the primitive processes for not only motor learning tasks, such as acquiring tool skills and adaptive movements, but also cognitive learning tasks such as language acquisition. We study fundamental algorithms and feasiblity of application about these processes for motor learning domains and cognitive learning domains.

  1. Megumi Miyashita, Ryo Hirotani, Shiro Yano, and Toshiyuki Kondo, Direct Policy Search with Extremum Seeking, SICE Annual Conference 2017, Kanazawa University, Japan. (9/22, 2017)

VR Rehabilitation

We developed an immersive VR system for analyzing how the visual interventions modulate bodily self-consciousness and effect on the neurofeedback training of motor imagery-based BCI. The immersive VR system enables an amputee to have strong bodily self-consciousness such as Sense of Agency (SoA) and Sense of Ownership (SoO), and it would be a promising intervention for reducing phantom limb pain.

  1. Shin Nagamine, Akira Ishii, Shiro Yano, Toshiyuki Kondo, Approach towards reduction of phantom limb pain using immersive virtual reality system, International Neurorehabilitation Symposium (INRS 2017), RehabWeek2017, London, (7/18-20, 2017)
  2. Shin Nagamine, Yoshikatsu Hayashi, Shiro Yano, Toshiyuki Kondo, An Immersive Virtual Reality System for Investigating Human Bodily Self-Consciousness, The 2016 Fifth ICT International Student Project Conference hosted by the Faculty of ICT, Mahidol University, Salaya Campus, Nakhon Pathom, Thailand (5/27, 2016)

BCI Neuro-rehabilitation

EEG-based Brain-Computer Interface for Neuro-rehabilitation

According to recent neuro-rehabilitation research, an appropriate re-afferent sensory feedback synchronized with a voluntary motor intention would be effective for promoting neural plasticity in stroke rehabilitation. Due to this, a BCI-based neuro-robotic rehabilitation is considered to be a promising approach. To detect the motor intention, an event-related desynchronization (ERD), which can be evoked by intrinsic motor imagery is usually used. However there exist various factors that affect ERD production, and its neural mechanism is still an open question. As a preliminary stage for realizing an effective neuro-robotic rehabilitation system, we evaluate mutual effects of extrinsic (visual and somatosensory stimuli) and intrinsic (spontaneous motor imagery) factors in ERD production. Experimental results indicate that these three factors are complicatedly interacting with each other and probably affect our sense of agency.

 System overview

  1. Toshiyuki Kondo, Midori Saeki, Yoshikatsu Hayashi, Kosei Nakayashiki, and Yohei Takata, Effect of instructive visual stimuli on neurofeedback training for motor imagery-based brain-computer interface, Human Movement Science, 2014, doi:10.1016/j.humov.2014.08.014. Journal site
  2. Kosei Nakayashiki, Midori Saeki, Yohei Takata, Yoshikatsu Hayashi and Toshiyuki Kondo, Modulation of event-related desynchronization during kinematic and kinetic hand movements, Journal of NueroEngineering and Rehabilitation, 2014, 11:9, DOI: 10.1186/1743-0003-11-90.
  3. Yohei Takata, Kotaro Takeda, Rieko Osu, Yohei Otaka, Toshiyuki Kondo, Koji Ito, A Proposal of EEG-FES based Rehabilitation System for Lower Limbs, The 26th Symposium on Biological and Physiological Engineering (BPES 2011), (9/22, 2011)

Human Motor Learning

Regarding simultaneous learning of two opposing force fi elds, it has been reported that a random schedule is significantly better than training that alternates at every trial, even if the total number of experiences is the same in both cases. Thus we assumed that this is because the alternating training schedule gave no opportunities of experiencing the same visuomotor rotation successively. To test this assumption, we compared three training conditions where the rotational transformation type (1) alternated every trial, (2) alternated every two trials, or (3) changed randomly. Experimental results suggest that providing subjects with the opportunity of successive trials has a small positive e ect on simultaneous learning of two con icting visuomotor rotations compared with training that alternates conditions at every trial. However, subjects trained with random schedule still show a signi cant advantage in comparison with subjects trained in either alternating schedule.
Also we developed a manipulandum that can realize arbitrary fore fields.

Motor Learning

  1. Takashi Sakamoto, and Toshiyuki Kondo, Visuomotor learning by passive motor experience, Frontiers in Human Neuroscience, 2015, doi: 10.3389/fnhum.2015.00279. Journal site
  2. Toshiyuki Kondo, Yuya Kobayashi, Takayuki Nozawa: Effect of Successive Experiences on Simultaneous Learning of Conflicting Visuomotor Rotations, IROS2008 Full Day Workshop, Nice, France, 26, September 2008.

EMG-based Cybernetic Interface

Foot Gestures

  1. Akira Ishii, Toshiyuki Kondo, and Shiro Yano, Improvement of EMG Pattern Recognition by Eliminating Posture-dependent Components, Proc. of the 14th International Conference on Intelligent Autonomous Systems (IAS-14), Shanghai, China (7/5, 2016)
  2. Yuma Sasaki, Toshiyuki Kondo, A Proposal of EMG-based Teleoperation Interface for Distance Mobility, The 2011 IEEE International Conference on Systems, Man, and Cybernetics (IEEE SMC 2011), Anchorage, Alaska (10/13, 2011)
  3. Chiharu Arakawa, Toshiyuki Kondo, A Study on Foot Gesture Recognition for Portable Device Operation, The 26th Symposium on Biological and Physiological Engineering (BPES 2011), (9/22, 2011)
  4. Toshiyuki Kondo, Osamu Amagi, Takayuki Nozawa: Proposal of Anticipatory Pattern Recognition for EMG Prosthetic Hand Control, Proc. the 2008 IEEE International Conference on Systems, Man, and Cybernetics (SMC 2008), Singapore, 12-15, October 2008.

Sports training support system using real time EMG feedback

Knowledge about skilled movements and its extraction techniques would be valuable for practical applications such as digital archiving of skilled technicians, sports training support systems for beginners, and developing a novel rehabilitation procedure for aged or disabled people. Aiming at the sports training support application, we have investigated human motor skills, such as basketball dribbling and throwing darts movement by using a high-speed camera and simultaneous recording of electromyogram.

Training System for Darts Throwing

  1. Tran Nguyen Bao, Shiro Yano, Toshiyuki Kondo, and Truong Quang Dang Khoa, Analyzing Effects of Variance in Kinematic Parameters on Performance and EMG in Dart Throwing, 2016 IEEE Sixth International Conference on Communications and Electronics (ICCE 2016), Novotel, Ha Long, Vietnam (7/28, 2016)
  2. Hiroshi Yamaguchi, Toshiyuki Kondo, Analysis of Motor Skill for Throwing Darts: Measurement of Release Timing, SICE Annual Conference 2011, Waseda University, Tokyo, (9/16, 2011)
  3. Seimei Abe, Takayuki Nozawa, Toshiyuki Kondo: A Proposal of EMG-based Training Support System for Basketball Dribbling, Proc. of HCI International 2009, San Diego, USA, 19-24, July 2009.

Personal robot design for a lasting human-agent interaction (HAI)

HAI experiments

  1. Takayuki Nozawa, Toshiyuki Kondo, Autonomous Adaptive Agent with Intrinsic Motivation for Sustainable HAI, Journal of Intelligent Learning Systems and Applications, 2, pp.167-178, 2010. Open Access
  2. Toshiyuki Kondo, Daisuke Hirakawa, Takayuki Nozawa: Sustainability and Predictability in a Lasting Human-Agent Interaction, Proc. the Eight International Conference on Intelligent Virtual Agents, pp.505-506, Tokyo, Japan, 2008.

Intelligent photo browser for flood of personal digital photographs

APC (Automatic Photo Classifier) demo:


  1. Yuki Orii, Takayuki Nozawa, Toshiyuki Kondo, Web-based Intelligent Photograph Management System Enhancing Browsing Experience, Journal of Advanced Computational Intelligence and Intelligent Informatics, Vol.14, No.4, pp.390-395, 2010.
  2. Yuki Orii, Takayuki Nozawa, Toshiyuki Kondo: User study of Automatic Photo Classifier by Color and Timestamp, Proc. of IEEE/WIC/ACM International Workshop on Web Intelligence (IWI’09), Milan, Italy, 15, September 2009.
  3. Yuki Orii, Takayuki Nozawa, Toshiyuki Kondo: Web-based Intelligent Photo Browser for Flood of Personal Digital Photographs, Proc. International Workshop on Web Intelligence (IWI’08), Sydney, 9-12, December 2008.

yesterday:3 today:1 total:6153