Abstract | ||
---|---|---|
Caregivers in nursing facilities are too busy to pay attention constantly to care receivers. Recently, some cameras have been set up in some nursing facilities. However, the caregivers cannot continuously monitor the care receivers on a display in the daytime. Now two-dimensional (2D) poses of many people in an image can be detected by OpenPose software; it can detect skeletons of humans by using a deep learning method. Therefore, we thought that if 2D poses of care receivers were detected by OpenPose immediately before a new action (preliminary action) and if the poses could be classified by a deep learning method, it might be possible to infer the care receivers' intentions. In this paper, we created a learning model that can discriminate the preliminary action based on coordinate data of keypoints detected by OpenPose from an image of a person. We examined whether a subject's action that was going to interrupt a conversation could be predicted or not. The result suggested that the learning model can discriminate the preliminary action of the subject by the coordinate data1.
|
Year | DOI | Venue |
---|---|---|
2019 | 10.1145/3319619.3326886 | GECCO |
Keywords | Field | DocType |
Conversation, Interruption | Computer science,Inference,Artificial intelligence,Machine learning | Conference |
ISBN | Citations | PageRank |
978-1-4503-6748-6 | 0 | 0.34 |
References | Authors | |
0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Ryuji Tanaka | 1 | 0 | 0.34 |
Chika Oshima | 2 | 27 | 9.13 |
Koichi Nakayama | 3 | 2 | 5.16 |