Abstract

It’s becoming increasingly easy for people to capture a short burst of images (or a video clip) instead of a simple still photograph. Image bursts and video clips are a rich source of information and can better capture the “moment”. However, their visual presentation is often more cluttered, and without careful planning and a lot of expertise, it is often very difficult for people to create visuals from image bursts and videos that are as compelling and easily consumed as still photographs.

In this talk, I will present recent work that explores the creation of “living” photographs known as “cliplets”, “cinemagraphs”, and video textures – a type of imagery that sits between videos and stills. Our work uses computer vision methods to automate aspects of the creation process such that an inexperienced user can create compelling imagery with only modest input.

Bio

Neel Joshi is a Researcher in the Graphics Group at Microsoft Research. His work is in computer vision, computer graphics, and (more recently) HCI, focusing particularly on imaging and computational photography applications, with various artistic endeavors squeezed in. Neel earned an Sc.B. from Brown University, an M.S. from Stanford University, and his Ph.D. in computer science from U.C. San Diego in 2008. He has held internships at Mitsubishi Electric Research Labs, Adobe Systems, and Microsoft Research, and he was a visiting professor at the University of Washington in 2010.

Abstract

For students, knowledge workers, and creative professionals alike, the vast tectonic shifts now taking place in the world of publishing, electronic books, device form-factors, and emerging modalities of natural interaction create tremendous opportunity to sunder new realms of innovation from the fabled Pangaea of reading.

These fault-lines are visible everywhere, if only one knows where to look. The moment a serious seeker of knowledge cracks open a book – electronic or otherwise – the fissures in the traditional monolithic view of reading begin to propagate. Multiple books and papers and yellow sticky-notes tile the tabletop like a riotous mosaic of stationary. Highlights emblazon the text and notations crowd the margins. These dedicated knowledge-seekers mark-up articles, tear out pages, and fill notebooks with handwritten insights.

And from this teeming jungle of information, they synthesize and create anew.

Natural ways of interacting with texts – from simple and expressive multi-touch gestures to markup with electronic pens to sensing the subtle motions of devices and their context of use – together promise, with appropriate design, to transform our experience of reading -- and thereby to transform our students and seekers into a newly empowered generation of creative professionals.

This talk will flash a few glimpses of this new world, this e-Creation, and demonstrate some concrete technologies and techniques that illuminate the way forward. It is a many-forked path fraught with design dilemmas and unknowns, but in my view the experience of reading and interacting with electronic information has only yet begun its incredible transformation.

Abstract

Low-latency teleoperation refers to highly interactive teleoperation in which human operators can control a remote robot in a temporally continuous way, much as they would control their arm or overall body
position. Depending upon the application, this control regime can also include control in which an alternative move-and-wait strategy might be used, provided that the operators do not engage in substantial strategic movement planning during the wait for the robot to catch up with their commands.

There are three factors that influence user difficulty during such low-latency operation: temporal, geometric, and procedural. The research to be discussed in this presentation primarily focusses on the consequences of the geometric misalignment of display and control axes, as occur when the remote camera used for teleoperation is misaligned with respect to the user's torso-referenced view. However, some studies of the interaction of latency and control difficulty due to the misalignment will be included since they demonstrate an interesting multiplicative interaction between these two factors.

We find that a constant-angular-targeting-error model may be used to model the trajectories of some of the targeting hand movements we have studied. This model is similar to an equiaxial spiral model discussed by Rushton and colleagues for human walking and by Tucker with respect to the diving trajectory of raptors,the bird of prey. These models point to a natural measure of the misalignment phenomenon yielding an explanation for the effects of rotations < ~65° about cardinal axes . However, deeper analysis will be needed to explain the effects of larger rotations. Our recent experimentation studying the most general type of 3D motions also reveals some interesting spatial anisotropies that probably reflect the way three dimensional movements are internally represented. In general the effect of the display control misalignment may be described by what we call the Misalignment Disturbance Function (MDF). Future research will be directed to explaining the specific features of this function.

Hand movement of a virtual blue spherical cursor to touch a virtual green spherical target under conditions of misalignment between display and control coordinates. This problem arises during teleoperation with the remote camera is not properly aligned.

Bio

Stephen R. Ellis has headed the Advanced Displays and Spatial Perception Laboratory at the NASA Ames Research Center. He received an A.B. in Behavioral Science from U.C. Berkeley (1969), a Ph.D. in Psychology (1974) from McGill University and has had postdoctoral fellowships in Physiological optics and Bioengineering at Brown University and at U.C. Berkeley respectively. He has published over 200 journal publications and reports on user interaction with spatial information and has been in the forefront of the introduction of perspective and 3D formats into user interfaces for aerospace applications. More recently he has pursued the study of virtual environments as user interfaces and as scientific instruments. He has served on the editorial boards of Presence, Virtual Reality and Human Factors and has edited a book, Pictorial communication in virtual and real environments, (Taylor and Francis, London, 2nd Ed., 1993).

June 27, 2012
Title: From Plastic to Pixels: In Pursuit of Effective Touch-Typing on Touch Screens

Speaker: Jacob O. Wobbrock, Information School, University of Washington

Abstract:
Fast, accurate, and satisfying text entry remains a significant challenge on touch screens. The lack of tactile feedback from physical keys and the loss of distinction between touching and pressing on touch screen keyboards are two of many challenges. The challenges increase when moving from large surfaces to mobile touch screens, where small screens and walking-induced situational impairments compromise accuracy.
In this talk, I will present a study of “touch-typing on flat glass” to understand finger-strike patterns for touch screen keyboards (CHI’11). I will also describe an adaptive keyboard built for Microsoft Surface that morphs its key layout to remain positioned beneath users’ fingers (CHI’12). I will also show a way to incorporate stroke gestures for non-alphanumeric input into this keyboard (CHI’12). For mobile text entry, I will describe WalkType, a keyboard that reduces a walker’s text entry error rate by incorporating touch features, accelerometer data, and inference about users’ walking behavior (CHI’12). Finally, I will describe Perkinput, a Perkins Brailler-based method for eyes-free text entry using Braille-like chord patterns (GI’12). Taken together, these projects highlight the potential for effective touch screen text input, and point to future possibilities where exciting work remains.

Bio:
Jacob O. Wobbrock is an Associate Professor in the Information School and an Adjunct Associate Professor in the Department of Computer Science & Engineering at the University of Washington, where he advises students from UW's information science and computer science programs. His research in human-computer interaction combines computer science, interaction design, and psychology to investigate novel user interface technologies, input and interaction techniques, human performance with computing systems, and accessible, mobile & surface computing interfaces. He has co-authored eight best paper winners and five best paper nominees, and is a recipient of an NSF CAREER award and three other NSF grants. He obtained his Ph.D. in Human-Computer Interaction from Carnegie Mellon University in 2006, and his B.S. and M.S. from Stanford University in 1998 and 2000, respectively. His cyberself inhabits a purple cave at faculty.washington.edu/wobbrock/.

June 20, 2012
Harshvardhan Vathsangam (USC)
Title: Sense and Sensibility: Statistical Techniques for Human Body Sensing
using Inertial Sensors.
Abstract:
Inertial sensors are everywhere, be it in our phones, video-game
controllers, exercise equipment, cameras and more. Their ever-reducing
size, plumetting costs and rapid miniaturization has led to their
widespread use in our daily lives. In this talk, I focus on two
applications of inertial sensors - in counting the number of calories
people burn everyday and in assisting communication for children with
cerebral palsy. Using machine learning techniques I show how data from
these sensors can be used in monitoring and improving quality of life.

Calorie tracking: Monitoring and tracking calories we burn forms an
important piece in promoting healthier lifestyles. Current techniques
that measure caloric consumption such as thermal chambers and
metabolic carts are expensive, cumbersome to use and are not portable.
In this talk we provide a family of algorithmic techniques to that
provide an accurate and "lcheap" mapping from one's movement to
calories consumed.

Alternative Communication: For a child with cerebral palsy learning
communication skills, typically involves one-on-one time with a
teacher learning how to use a alternative augmentative communication
device. However, if the child has limited communication skills the
he/she has the additional burden of simultaneously learning how to use
a device and learning to communicate. We propose to mitigate this
problem using an inertial sensor based device that interactively
learns a dictionary of the child's movements. These are then mapped to
a basic vocabulary and form an initial bridge to more advanced
communicative ability.

Bio:
I am a Ph.D. candidate from the Robotic Embedded Systems Laboratory,
Viterbi School of Engineering, University of Southern Califoria. My
advisor is Prof. Gaurav S. Sukhatme. I hold an undergraduate degree in
Engineering Physics from IIT, Madras.

My primary area of research is in statistical machine learning for
clinical healthcare monitoring and ubiquitous sensing with mobile
phones. I also work on design of on-body inertial sensor hardware,
machine learning for Kinect-based interfaces and alternative
augmentative communication devices for children with cerebral palsy.
Website: robotics.usc.edu/~harsh