RECG |
uWave: Accelerometer-based Gesture Recognition
|
||||||||
|
uWave is a gesture recognizer based on a 3-D accelerometer. It was a project held by Rice Efficient Computing Group (RECG) of Rice University in collaboration with Motorola Labs in 2007/08. I became a research assistant of the group when I was an exchange student in Rice University and I took charge of the uWave project. We did not use Hidden Markov Model (HMM) popular in speech and some gesture recognitions as it requires a large amount of training samples, which often causes inconvenience in human-computer interaction. Dynamic Time Warping (DTW) algorithm and template adaptation were applied on the recognizer. It achieves 98.6% accuracy and only requires single training sample. It also allows users to employ personalized gestures. I was responsible for the design and development of the DTW algorithm on the gesture recognizer and part of the user study on the accuracy testing and improvement. Conference and Journal Articles
Extended Abstract and Technical Report
Gesture Library Acceleration data of 4480 gesture samples are collected from eight
participants for seven days. More details can be found in our PerCom paper. Related Applications' Source Code
The above 3 applications call uWave.h and uWave.c Demo Video on YouTube You can also watch the demonstration on YouTube: http://www.youtube.com/watch?v=rrZfLAfeZww (View this page in Romanian courtesy of azoft, in Indonesian courtesy of ChameleonJohn, in Portuguese courtesy of Travel Ticker, and in Macedonian courtesy of Jim Anastasovski.) |
||||||||
[Home][People] [Publications] [Projects] [Downloads] [News] [Acknowledgments] [Contact us] |