We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
A selective region-based detection and tracking approach towards the recognition of dynamic bare hand gesture using deep neural network.
- Authors
Yadav, Kuldeep Singh; Anish Monsley, K.; Laskar, Rabul Hussain; Misra, Songhita; Bhuyan, M. K.; Khan, Taimoor
- Abstract
Complexities such as variations in pose, scale, speed, illumination, occlusion, etc., in detecting the gesture object at the first frame and tracking it in a dynamic environment make it challenging for gesture-based human–computer interactions. The kernel-based operation on the first frame requires additional computational resources for detecting the bare hand. Therefore, a region-based detection approach is proposed to detect the bare hand-like regions using skin color and movement information. Using transfer learning, the AlexNet is used for detecting the bare hand from this region of interest. To track it in the successive frames, a Detection and Tracking (DaT) algorithm is proposed. The smoothen trajectory is utilized for training a separate deep neural network to recognize 60 isolated dynamic gestures. An average accuracy of 97.80%, 96.20% is achieved for the bare hand detection on the Oxford hand database, NITS hand gesture VIII database. The experimental observation reveals that the proposed method is computationally efficient as compared to the existing models. For the recognition of 60 gestures, an average accuracy of 96.48% is achieved. The proposed system provides a relative improvement of 17% over the base models.
- Subjects
HUMAN skin color; GESTURE; HUMAN-computer interaction; OBJECT recognition (Computer vision)
- Publication
Multimedia Systems, 2022, Vol 28, Issue 3, p861
- ISSN
0942-4962
- Publication type
Article
- DOI
10.1007/s00530-022-00890-1