Comments
Reference
A $3 Gesture Recognizer - Simple Gesture Recognition for Devices Equipped with 3D Acceleration Sensors
Sven Kratz, Michael Rohs
IUI 2010 - Hong Kong
Summary
This paper presents an algorithm for using 3D accelerometer data in the task of gesture recognition. The algorithm presented is simple in nature and easily implementable for prototyping situations. It uses only trigonometric and geometric calculations and is sensor independent. It has a success rate of 80% and can support about 10 to 15 gestures successfully. The recognizer requires little training data. A user study with 12 participants indicated the 80% success rate.
Analysis
I was greatly disappointed to find out that this was not a $3 hardware device but rather an algorithm. However, I really appreciated the direct and thorough nature of this paper. It presented its exact claims and its shortcomings. It acknowledged its scope of application and its reason for being a good choice within that domain. Additionally, it’s always nice to have some basic algorithms to use when you aren’t looking to be the world’s best at something (like when prototyping). Overall this paper illustrated the power of simple algorithms and their usefulness.
It is always nice to read not only about the good things about a development, but also their shortcomings. When I read this paper, understanding their results was a big part of the paper, since they clearly state that even though their results cannot compare to other algorithms out there, for the simplification they had added, it was a good development.
ReplyDelete