PhysicsBook: A Sketch-Based Interface for Animating Physics Diagrams
IUI 2012, February 2012, Lisbon, Portugal
Salman Cheema
· University of Central Florida
· Pursuing a Doctoral degree in Computer Science
· Area of primary interest is Computer Graphics.
· http://www.eecs.ucf.edu/isuelab/people/salman.php
Joseph J. LaViola Jr.
· University of Central Florida
· “interests include pen-based computing, 3D user interfaces for games, human motion estimation, virtual reality, and interactive computer graphics”
· http://www.eecs.ucf.edu/~jjl/
Summary
The current method of solving physics problems lacks
feedback. PhysicsBook is designed to
provide that feedback to new students and instructors. “We envision it to have the following feature
set:” natural interaction, robust recognition, reasoning, feedback, and
seamless animation.
There are several other forms of sketch recognition, but
“none of these applications have intelligent tutoring in physics as their
motivation.” Other researchers are
working on making a sketch-based tutoring system but none of those other programs
allows the user to draw images. The
current design is limited to basic physics concepts. A major advantage over the previous attempts
is “the use of data transformations that allow [PhysicsBook] to deal with
instances where the given problem solution cannot be directly used for
animation.”
The core components need to do three main things:
recognitions, inference and animation. An
important aspect is “the inference of intent in ambiguous cases.” Users can make associations between
mathematical expressions and recognized sketch primitives. “PhysicsBook can
deal with equations including both numeric and symbolic parameters.” “A ‘Reset’ mode is provided that lets users
debug and correct existing equations and associations.”
The first step to creating a diagram is the sketch
recognition. The items that can be
recognized are: circles, polygons, springs, wires, and pulleys. Composite primitives are created when two or
more objects are overlapping one another on the page. Single primitives are recognized before the
composite primitives. The inference
subsystem takes recognized primitives and annotations and infers associations
between them as well as establishing initial values. “PhysicsBook uses a
customized 2D physics engine.” “Users
can specify their own equations for forces, velocity, position, and acceleration
directly which can augment or even replace the standard update mechanism of the
physics engine.” Users can animate
problems from a range of physics domains related to classical mechanics. Concepts that are only indirectly related to f=ma
must be transformed into an acceptable input before
animation can occur. For transformed
variables, the results will be magnitude only, so user inputs are needed to
determine the direction.
 |
| Figure 1. |
There was a small scale user study to get preliminary
feedback. Five participants from UCF
were recruited for an informal evaluation of PhysicsBook. They were given an introduction to the use of
PhysicsBook. The users wrote out the
solutions to a physics problem and then filled out a 7 point Likert scale questionnaire. Overall the results were very positive with
two exceptions: drawing of springs and recognition of mathematics. The participants in our study preferred to
get real-time feedback about the handwritten mathematics. An example of a simple physics question is
shown in Figure 1.
“Currently the capability of PhysicsBook to animate a given
problem is limited to select cases in domains related to classical mechanics.” The foundation created with PhysicsBook will
be advantageous in providing natural interfaces for understanding physics.
Related Work
1.
“HMM-based efficient sketch
recognition” – This article explains that sketch recognition is an incremental
process, much like in PhysicsBook, and explains the running time for their
algorithm.
2.
“SketchREAD: a multi-domain sketch
recognition engine” – Another article on an alternative way to have sketch
recognition and discusses some of the common pitfalls.
3.
“Hierarchical parsing and
recognition of hand-sketched diagrams” – This paper presents an “integrated sketch parsing and recognition approach designed
to enable natural, fluid, sketch-based computer interaction.”
4.
“A toolkit approach to sketched diagram
recognition” – This paper covers an implementation for a sketch tool framework
instead of creating new software for sketch recognition.
5.
“An image-based, trainable symbol recognizer for hand-drawn sketches” – This
article describes a symbol recognizer instead of a sketch recognizer.
6.
“A Parsing Technique for Sketch Recognition
Systems” – This paper describes a framework for generating parsers and modeling
sketch languages.
7.
“Polygon recognition in sketch-based interfaces with
immediate and continuous feedback” – This paper details taking a sketch and
creating the perfect geometric representation in real-time, which PhysicsBook
can do, but PhysicsBook also can work after completion.
8.
“A retargetable framework for interactive
diagram recognition” – This paper describes a “retargetable framework which can be
used to speed the development of robust interactive sketch recognition systems.”
9.
“Sketch-based modeling with few
strokes” – This paper covers simple ways of creating 3D objects with just a few
strokes.
10.
“A sketch-based interface for
iterative design and analysis of 3D objects” – This paper describes a program
that works in conjunction with CAD to make freehand sketch based engineering
design.
As is clearly obvious from the above listed papers, this topic is
not novel. The only novel thing about it
is to apply the idea of sketch recognition to a tutoring program for physics.
Evaluation
The initial creation of the PhysicsBook program required the
authors to evaluate the results in a quantitative, objective manner. If any of the answers had been incorrect, the
whole result was incorrect. There is no
variation allowed in those results. On
the other hand, the results from the questionnaire are quantitative and
subjective. The use of a 7 point Likert
scale allow the users to choose how effective the program feels to them without
too much variety and the numbers can be averaged to provide general results as
they were in this paper. Overall, the
evaluation was effective because it allows the users to have their input since
no two users will have the same ideas.
Discussion
I think that this program is a fantastic idea; if this
program had been around when I was taking physics classes I would have been
able to learn the material much faster.
The evaluation for this paper was highly effective; it allows the users
to show the effectiveness of the program and how they thought of the program
with quantifiable data that can be averaged.
There is not a better way to test a new program, subjective values are
needed but quantitative values are best for ease of calculations. This contribution is not entirely novel,
there are clearly other people who have worked on areas extremely similar to
the ones discussed here. Though this
work is not entirely novel, there are parts of this program which are new and
combining all of the programs into one, like PhysicsBook, is a wonderful and
practical idea.
No comments:
Post a Comment