Information Technology Reference
In-Depth Information
Dynamic Time Warping-Based K-Means
Clustering for Accelerometer-Based Handwriting
Recognition
Minsu Jang 1 , Mun-Sung Han 1 , Jae-hong Kim 1 , and Hyun-Seung Yang 2
1 Electronics and Telecommunications Research Institute
Gajeong-dong 161, Yuseong-gu, Daejeon-si, 305-700, South Korea
{ minsu,msh,jhkim504 } @etri.re.kr
2 Korea Advanced Institute of Science and Technology
291 Daehak-ro, Yuseong-gu, Daejeon-si, 305-701, South Korea
hsyang@kaist.ac.kr
Abstract. Dynamic time warping(DTW) is widely used for accelero-
meter-based gesture recognition. The basic learning strategy applied with
DTW in most cases is instance-based learning, where all the feature
vectors extracted from labeled training patterns are stored as reference
patterns for pattern matching. With the brute-force instance-based learn-
ing, the number of reference patterns for a class increases easily to a
big number. A smart strategy for generating a small number of good
reference patterns is needed. We propose to use DTW-based K-Means
clustering algorithm for the purpose. Initial training is performed by
brute-force instance-based learning, and then we apply the clustering
algorithm over the reference patterns per class so that each class is rep-
resented by 5 10 reference patterns each of which corresponds to the
cluster centroid. Experiments were performed on 5200 sample patterns
of 26 English uppercase alphabets collected from 40 personals using a
handheld device having a 3-d accelerometer inside. Results showed that
reducing the number of reference patterns by more than 90% decreased
the recognition rate only by 5%, while obtaining more than 10-times
faster classification speed.
Keywords: Dynamic Time Warping, K-Means, Accelerometer, Gesture
Recognition.
1
Introduction
Nowadays, most mobile devices, including game console controllers and smart
mobile phones, are designed to include inertial sensors for improved natural user
experience. The most popular inertial sensor is the MEMS-based accelerometer.
With accelerometer, it is possible to capture and recognize the posture as well as
the motion of devices. Several research efforts tried to develop gesture recognition
systems based on acceleration signals from mobile devices [3-5].
In many of such efforts, template matching-based 1-nearest neighbor classifica-
tion combined with dynamic time warping(DTW) has been successfully applied,
 
Search WWH ::




Custom Search