BROWSE
Detail
Conference Paper
3D Articulated Human Pose Recognition via Learning Deep Gaussian Mixture Models for Virtual Exercise
Cited 0 time in 
Share 
Share
- Authors
- Jong-Sung Kim
- Issue Date
- 2023-10
- Citation
- International Conference on Consumer Electronics (ICCE) 2023 : Asia, pp.79-81
- Publisher
- IEEE
- Language
- English
- Type
- Conference Paper
- Abstract
- In this paper, a new 3D articulated human pose recognition method based on deep Gaussian mixture models is proposed for a virtual exercise game using a RGB-D camera. The proposed method acquires 3D skeleton data for each human pose with the RGB-D camera. Then, a new deep Gaussian mixture model (DGMM) is learned from 3D skeleton data for each human pose. The proposed method can recognize each 3D articulated human pose by applying the DGMM network to 3D skeleton data stream of the RGB-D camera. The performance of the proposed method is verified with real experiments.
- KSP Keywords
- 3D skeleton, Data stream, Gaussian Mixture Models(GMM), Gaussian mixture(GM), Human Pose Recognition(HPR), RGB-D cameras, Recognition method, Skeleton data, real experiments