발간년도 : [2021]
| 논문정보 |
|
| 논문명(한글) |
[Vol.16, No.5] Validation and Analysis of Human Behavior Database Data for Robot Learning |
|
| 논문투고자 |
Byung-Tae Chun |
|
| 논문내용 |
A social robot is a robot that interacts with people through social actions such as language and gestures. Recently, research on human behavior recognition of social robots is being actively conducted. Human behavior recognition refers to a robot that recognizes and judges human motions and movements. The application fields of social robots are used in various fields such as human care, services, and conversation partners. An experimental data set is needed to study human behavior recognition of social robots. The data set for behavior recognition mainly consists of behaviors for human-robot interaction or daily life behaviors. Recently, public data sets are being actively disclosed, and if necessary, behaviors necessary for recognition are individually constructed and used. For human recognition, the video data set performs tagging that classifies actions into sections. Recently, neural network methods have been widely used to recognize human behavior of robots. Neural network learning basically proceeds under the assumption that tagging is correct. However, if the timing of tagging the dataset used for neural network recognition is wrong, it can seriously affect neural network learning. Existing studies do not mention the problems that occur during tagging. In this paper, we try to evaluate the quality of the database built through tagging data analysis. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
