BROWSE
Detail
Conference Paper
Infant abnormal behavior classification through weakly supervised learning
Cited - time in 
Share 
Share
- Authors
- YEJIN LEE, Kyekyung Kim, Jaehong Kim, Seung-Ik Lee
- Issue Date
- 2020-10
- Citation
- International Conference on Intelligent Robots and Systems (IROS) 2020, pp.12118-12121
- Publisher
- IEEE
- Language
- English
- Type
- Conference Paper
- Abstract
- Nowadays, caregivers are focusing on the health information of newborn infants. It is very important to discover a neurological disease in infants. Especially, Early diagnosis must be made, in the case of premature infants as the possibility of neurological diseases is high. There are many systems for monitoring infants, but few systems classify the status of infants through artificial intelligence analysis. In this paper, the normal and abnormal movements were classified to help diagnose neurological diseases of infants by using Fully Convolutional Network (FCN) according to the skeleton information of infants. Besides, only the labels of normal and abnormal can produce the following weakly-supervised three characteristics. 1) whether the infant's condition is normal or abnormal) 2) which segment of the video shows the infant's movement is abnormal, 3) which body part of the infant is found to be abnormal. The proposed network is simple and suitable for application to robot devices that observe infants in the home environment. This study was conducted using the skeleton database extracted from the video of infants.
- KSP Keywords
- Abnormal behavior, Behavior classification, Early diagnosis, Fully Convolutional Networks(FCN), Home Environment, Intelligence Analysis, Neurological disease, Premature infants, Skeleton information, Weakly supervised learning, artificial intelligence