Paper Reading #19: A $3 Gesture Recognizer

Comments:

Shena Hoffman

Joe Cabrera

Reference Information:
Title: A $3 Gesture Recognizer - Simple Gesture Recognition for Devices Equipped with 3D Acceleration Sensors

Authors: Sven Krats and Michael Rohs

Presentation: (Conference Paper) IUI'10, February 7-10, 2010, Hong Kong, China

Summary: In this paper, the authors present a very simple  gesture recognizer for input devices with 3D acceleration sensors, such as a Nintendo Wii control.  Their development is really simple, not complex to use or implement. Thus, it is a great system to use for testing prototypes. The authors based much of they research and design on the "$1 Recognizer" by Wobbrock, a 2D gesture recognizer. Their recognizer is an extension of it, but also a really simple one. The main advantage and major contribution of this system is their "true" ability to recognize 3D motion.  

Figure 1. The reference gesture vocabulary containing
the gesture classes used for the preliminary evaluation.
(b) describes a clockwise circular motion, (c) a wrist
rolling motion (e) stands for a gesture resembling the
serve of a tennis player and (j) represents a repeated
rapid forward-backwards motion.

They evaluated the system with a user study with twelve participants and a set of 10 unique gesture classes.The study consisted in having each participants to enter a gesture class fifteen times using a WiiMote. Their results demonstrate an 80% recognition rate. However, the individual recognition rate varied from 58% to 98%. The authors do state that their recognition rate is lower than  than that showed in previous works. But, they believe this is an expected rate since they utilized simpler methods and implementing more gesture classes but less gesture training per class.

Discussion: I don't really know much about this type of recognizers, so I wouln't be certain if there is any other developments supporting 3D gesture recognition. Even though it seems like a very simple development, and the results are not very high or better than previous works, this may just be a start. I like the idea of them finding a use for their system. Using this for prototype testing sounds like a reasonable application for it.