Action recognition with adaptive RBFNN

Author

Aphaipanan, Srisuda ; Kidjaidure, Yuttana

Author_Institution

Dept. of Electron., King Mongkut´s Inst. of Technol. Ladkrabang (KMITL), Bangkok, Thailand

fYear

2014

fDate

5-8 March 2014

Firstpage

1

Lastpage

5

Abstract

This paper presents a method for action recognition by Adaptive Radial Basis Function Neural Network (ARBFNN) based on 3 dimensional human models. Recently, the action recognition of human is popular for the interactive applications caused many researchers tried to develop the algorithm and to find the features that have high performance. So this paper employed the features from the scalar part of Quaternion rotation that uses lower dimension than the conventional Cartesian features. Also, the Fuzzy C Means technique was used for pre-training the Radial Basis Function Neural Network (RBFNN). This method was tested with the CMU MoCap database and showed high recognition rates with small computation time.

Keywords

fuzzy set theory; image motion analysis; image recognition; learning (artificial intelligence); radial basis function networks; 3 dimensional human models; ARBFNN; CMU MoCap database; RBFNN pretraining; action recognition; adaptive RBFNN; adaptive radial basis function neural network; fuzzy C means technique; interactive applications; quaternion rotation; radial basis function neural network pretraining; Classification algorithms; Clustering algorithms; Hidden Markov models; Joints; Quaternions; Three-dimensional displays; Vectors; Fuzzy C Means; Quaternion; Radial Basis Function; pre-training;

fLanguage

English

Publisher

ieee

Conference_Titel

Information and Communication Technology, Electronic and Electrical Engineering (JICTEE), 2014 4th Joint International Conference on

Conference_Location

Chiang Rai

Print_ISBN

978-1-4799-3854-4

Type

conf

DOI

10.1109/JICTEE.2014.6804095

Filename

6804095