Bodily posture recognition with weighted dimension on kinect data stream

Conference proceedings article

Authors/Editors

Strategic Research Themes

No matching items found.

Publication Details

Author list: Jariyavajee C., Sirinaovakul B., Polvichai J.

Publisher: Springer

Publication year: 2018

Volume number: 566

Start page: 150

End page: 159

Number of pages: 10

ISBN: 9783319606620

ISSN: 2194-5357

eISSN: 2194-5357

URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85022212268&doi=10.1007%2f978-3-319-60663-7_14&partnerID=40&md5=80d55b230f875d8657d6d9bd1495495c

Languages: English-Great Britain (EN-GB)

View on publisher site

Abstract

The characteristic of the data stream is continuous, non-stationary, and very large or infinite size. Data stream classification requires the algorithm that able to classify data instance and learn from data incrementally. In this paper, the algorithm with Weighted Dimension is proposed and applied for the Kinect bodily posture recognition. The human body portions, as the input features, are calculated from Skeleton Joint data. The proposed algorithm successes in recognizing three human postures: stand, sit_on_chair, and sit_on_floor. The result of classification is 99.02% on average and 100% on moving accuracy. Moreover, the algorithm always learns from the data instances and some labels so the algorithm is able to learn whether the data instances are changed. In the other words, the algorithm could handle the concept drift in the data stream. ฉ Springer International Publishing AG 2018.

Keywords

Concept drift, Data stream classification, Kinect, Posture recognition