The Community for Technology Leaders
2016 International Conference on Big Data and Smart Computing (BigComp) (2016)
Hong Kong, China
Jan. 18, 2016 to Jan. 20, 2016
ISSN: 2375-9356
ISBN: 978-1-4673-8795-8
pp: 423-426
Hyun-Gook Kang , Knowledge Service Engineering, KAIST, Korea
Sang-Hyun Lee , Electrical Engineering, KAIST, Korea
ABSTRACT
This study proposes a technique to generate effective features to classify fundamental human body postures in image sequences such as standing, sitting on the chair, sitting on the floor, bending, and lying down. Truncated discrete cosine transform (DCT) is utilized to obtain features before performing truncated singular value decomposition (SVD). It has been shown that the truncated DCT disregards unnecessary values and thus makes features more simple and light, resulting in an improvement in classification speed. Moreover, this study verifies that the newly extracted features contribute to an increase in the accuracy of the human posture classification, and a definite decrease in distinction errors for bending and sitting postures.
INDEX TERMS
Feature extraction, Discrete cosine transforms, Surveillance, Image sequences, Mathematical model, Singular value decomposition, Cameras
CITATION

Hyun-Gook Kang and Sang-Hyun Lee, "Human body posture recognition with discrete cosine transform," 2016 International Conference on Big Data and Smart Computing (BigComp)(BIGCOMP), Hong Kong, China, 2016, pp. 423-426.
doi:10.1109/BIGCOMP.2016.7425962
94 ms
(Ver 3.3 (11022016))