Prof. Michael Fairhurst
Title: Is handwriting still relevant in the digital age?
Abstract: I have chosen a possibly rather provocative title for my presentation, since there is no doubt that our view of handwriting, its status, its utility and its necessity, has changed significantly in recent years. As a medium for personal communication, handwriting can no longer be automatically considered the modality of choice, and some might even claim that it has no serious long-term future at all. So, is handwriting destined to become practically useful only for the technology-deprived or the information-poor, or should we rather be looking towards a new age of handwriting pre-eminence?
These are interesting questions, but it may be difficult to answer them directly. My aim in this presentation is to consider the versatility of handwriting, and show how very different handwriting-oriented scenarios can offer very different perspectives on how we think about the fundamental nature of handwriting, and how we define its functionality and value. If we think of handwriting only in terms of its capabilities as a communication medium, then we are missing the richness and variety of its possibilities, and simultaneously failing to address some important technological opportunities too. Handwriting analysis clearly has implications way beyond its obvious value in supporting personal communication, therefore, but how should technology develop in order to address future needs?
Thus, the core of the presentation will aim to reassess the value of handwriting in the changing landscape of societal and technological developments. This will allow us to explore the many different facets of handwriting and drawing, and how we might exploit human ability for handwritten expression in a range of application domains covering communication, biometrics and security, in forensic analysis, and even in healthcare and medical diagnostics. We will see how some common threads run through very different application scenarios – for example, in work on the predictive properties of handwritten information – and how technological solutions to current problems can actually also create new exploitation opportunities.
Hence, the presentation will start from some general observations, then focus around some specific examples of technology application, and conclude with some crystal ball gazing, asking whether and how handwriting can retain its credibility and primary role in human activity in the future.
Biography: Michael Fairhurst is Professor of Computer Vision at the University of Kent in the UK, where he was Head of Department for 5 years until 2008. His research interests focus on computational architectures and algorithms for image analysis and classification, and applications including handwriting analysis and document processing, medical image analysis and security/biometrics. Current research includes work on multimodal biometrics, on the effect of ageing on biometric system reliability and on the analysis of handwriting, both for identification purposes, and in forensic applications. Biometric processing also underpins work which is investigating document encryption linked to biometric data. In particular, he is further developing work he has pioneered to establish novel techniques for the assessment and monitoring of neuropsychological conditions through the analysis of writing and drawing. Professor Fairhurst has been a member of numerous Conference (for example, ICDAR, ICFHR, IGS), Workshop, Academic, Technical and Government Committees (for example, as a member of the UK Government’s Biometrics Assurance Group), is an Editorial Board member of several international Journals, and is the Editor-in-Chief of the recently launched IET Biometrics Journal. He is also a member of the international committee overseeing the IEEE Certified Biometrics Professional (CBP) programme. He has published some 350 papers in the scientific literature, and is an elected Fellow of the International Association for Pattern Recognition (IAPR) in recognition of his contributions to the field, currently serving on the IAPR Education Committee.
Prof. Seiichi Uchida
Title: Dynamic time warping for comparing temporal handwriting trajectories and its recent extensions.
Abstract: Comparing two temporal handwriting trajectories is a fundamental task of online character recognition, writer identification, and handwriting analysis. Dynamic time warping (DTW), or DP matching, has been utilized for this task from 1970s. One main function of DTW is to provide the optimal nonlinear temporal correspondence between two trajectories. The nonlinearity is useful for dealing with nonlinear temporal fluctuation during handwriting. Another function is to evaluate the similarity or dissimilarity between two handwriting trajectories. Different from Euclidean distance, DTW can provide a similarity even when two trajectories have different lengths.
In this talk, I will introduce several extensions of DTW, which might be useful for graphonomics. The first extension is asynchronous DTW. In the conventional DTW, x and y elements representing pen-tip position at a time t are coupled tightly during DTW. In contrast, in the asynchronous DTW allows to couple x and y elements from different timings. This asynchronous usage of the x and y elements can be considered as a new deformation model of handwritings. The second extension is non-Markovian DTW. Usually, DTW and its stochastic version, HMM, assume the Markovian property that a pen-tip position (or other features) at a time t depends only on that at t-1. However, if we consider the writing process of “0”, this assumption is not accurate; in fact, we have to be careful that the first and the last pen-positions should be close enough for creating “0” as a circle and thus our writing process is highly non-Markovian. I will explain that a non-Markovian DTW can be realized by utilizing a graph-cut optimization scheme instead of a traditional dynamic programming optimization scheme. Other extensions, such as logical DTW, non-uniform DTW, and permutation-free DTW (called cube-search), will also be explained.
Biography: Prof. Seiichi Uchida is a professor of Kyushu University, Fukuoka, Japan. His research interests include character recognition, pattern recognition theories and image processing. He received 2007 IAPR/ICDAR Best Paper Award and 2010 ICFHR Best Paper Award in addition to several domestic awards, such as 2002 IEICE PRMU Research Encouraging Award, 2008 IEICE Best Paper Award, MIRU2006 Nagao Award (best paper award), MIRU2011 Excellent Paper Award. He has co-authored about 50 journal papers and 100 international peer-reviewed conference papers in this area. He is an Editorial Board member of IJDAR, General Chair of CBDAR2009 and CJKPR2010, PC Co-Chair of DAS2012, an area chair of ICPR2012, and a regular PC member of various important conferences, ICPR, ICDAR, ICFHR, DAS, ACCV, SSSPR, etc.
Prof. Hiroshi Yokoi
Title: Robotics rehabilitation projects based on mutual adaptation.
Abstract: Prosthetics and Rehabilitations for the handicapped person requires new and reliable robotics technology. This study investigates the reactions of our brain to an adaptable robotics technologies for handicapped person by using adaptable computation for biological signal and robot devices with large degrees of freedom. The proposed technologies are structured EMG (Electromyogram) – controlled robot hand and power assist devices with a learning function for EMG pattern recognition of transradial (below elbow) prostheses. It can be applied to amputees in the age group of 1 to 60. The key topic of this study is the mutual adaptation of a human and an adaptable robot hand. This is analyzed by using fMRI to clarify the plasticity of the motor and sensory areas of the cortex with a change in prosthesis. And also those devices obtained the motor function assisting for paralyzed person based on biofeedback by electrical stimulation through surface electrodes. The results are stable characteristics of hand grasping and walking assist system, and high discrimination reliability of intentions of motion. The mechanical design is structured by interference driven mechanism in order to adapt for many size of human hand/finger. The biofeedback device is designed by using the electric stimulator which works for wide range stimulation from -90V to +90V, and produces phantom sensing to show complex image of muscle contraction. This device is now started to applying for motor function assistance of stroke person. The presentation will show experimental results of prosthetics and rehabilitation with brain plasticity.
Biography: Dr. Hiroshi Yokoi was born in Nagoya, Japan, in 1963. He received B.S., M.S., and Ph.D. degrees in Precision Engineering from Hokkaido University, Japan, in 1986, 1990, and 1993, respectively. He was an engineer at Toyota Motor Cooporation from 1996 to 1997. He joined as the Researcher, Institute of Bioscience and Human Technology, AIST form 1993-1995. He was an Associate Professor in Department of Complex System Engineering of Faculty of Engineering at Graduate School of Hokkaido University from 1995 to 2004, and Department of Precision Engineering, Faculty of Engineering, The University of Tokyo from 2004 to 2009. He is a Senior Researcher in AI-Lab at the University of Zurich, from 2002 and also a Fellow Researcher in IAS-Lab at the University of West England form 2003, and is working as a Guest Professor, National Institute of Physiological Science from 2009. He is currently Professor of Department of Mechanics and Intelligence at the University of Electro-Communication, and Guest Professor of Interfaculty Initiative in Information at The University of Tokyo. His current research interests include computational intelligence in robotics, artificial life, medical engineering.