Human Sitting-Posture Recognition Based on the Cascade of Feature Mapping Nodes Broad Learning System
In order to effectively recognize the sitting posture of human in the office, the cascade of feature mapping nodes broad learning system is proposed. The Kinect is used to obtain the relevant data and to establish the sittingposture recognition database. The sitting-posture recognition module is inn...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Editorial Department of Journal of Nantong University (Natural Science Edition)
2020-09-01
|
| Series: | Nantong Daxue xuebao. Ziran kexue ban |
| Subjects: | |
| Online Access: | https://ngzke.cbpt.cnki.net/portal/journal/portal/client/paper/9c7d43d8e7b8dfa0e69aa36c7ecd41fc |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In order to effectively recognize the sitting posture of human in the office, the cascade of feature mapping nodes broad learning system is proposed. The Kinect is used to obtain the relevant data and to establish the sittingposture recognition database. The sitting-posture recognition module is innovatively designed based on the cascade of feature mapping nodes. By cascading feature mapping nodes, the low-level features are effectively mapped to highlevel features, and the discrimination of features is improved which can conveniently recognize the different sitting postures. Since the real videos contain tween frames between transformations on different sitting postures, the discriminant probability of frame and the structural similarity index are introduced to establish a tween frame detection module in the video sequences, which can select tween frames and improve the recognition accuracy. The experimental results in public datasets and self-built dataset show that the model not only achieves better performance on public datasets, but also has the average recognition accuracy of 99.90% on images and 79.21% on videos, which is 5.5%higher than the classic broad learning system. Its recognition accuracy is obviously improved, the speed is increased,and it has good performance on the generalization. |
|---|---|
| ISSN: | 1673-2340 |