Steven Hennessy
Patrick Frith
Reference Information
Title: A $3 Gesture Recognizer - Simple Gesture Recognition for Devices Equipped with 3D Acceleration Sensors
Authors: Sven Kratz, Michael Rohs
When/Where: IUI 2010
Authors: Sven Kratz, Michael Rohs
When/Where: IUI 2010
Summary
This paper discussed a simple gesture recognizer using 3D accelerometer input from a device such as a mobile phone or Wii controller. The algorithm uses only trigonometric and geometric calculations, and is designed to be device-independent without the use of special tool-kits or frameworks. About 5 training examples are required to perform the recognition. The approach is quick, cheap, requires minimal training, and a decent recognition rate.
Testing of the system was done using a Wii controller. Each participant in the evaluation study was asked to complete a set of gestures, entering each one 15 times. The first 5 were used for training, and the rest for evaluation of recognition. The evaluation showed that correct recognition was lower than that of a more complex system, but for the purposes of this system (quick responses for prototyping systems) it is acceptable.
Discussion
This paper discussed a simple gesture recognizer using 3D accelerometer input from a device such as a mobile phone or Wii controller. The algorithm uses only trigonometric and geometric calculations, and is designed to be device-independent without the use of special tool-kits or frameworks. About 5 training examples are required to perform the recognition. The approach is quick, cheap, requires minimal training, and a decent recognition rate.
Testing of the system was done using a Wii controller. Each participant in the evaluation study was asked to complete a set of gestures, entering each one 15 times. The first 5 were used for training, and the rest for evaluation of recognition. The evaluation showed that correct recognition was lower than that of a more complex system, but for the purposes of this system (quick responses for prototyping systems) it is acceptable.
 |
| The set of gestures used for evaluation. Source: paper. |
I thought this was a good idea, since gesture recognition generally takes a lot of computation. However, it seems like there were a lot of drawbacks of this system. For instance, the gesture vocabulary could only have 10-15 gestures before the computational overhead became too large to be practical. So while this system may be useful for prototype systems that need fast feedback, it seems like it would be rather useless in a complete system. With that said, I think it could be useful to continue to investigate faster and different methods for gesture recognition, since more and more devices seem to be providing uses for such a system.
Definitely a system for prototyping and quick use needs. I do believe they could expand it somewhat while keeping it simple though.
ReplyDelete