In the last few years, human action recognition potential applications have been studied in many fields such as robotics, human computer interaction, and video surveillance systems and it has been evaluated as an active research area. This paper presents a recognition system using deep learning to recognize and identify human actions from video input. The proposed system has been fine-tuned by partial training and dropout of the classification layer of Alexnet and replacing it by another one that use SVM. The performance of the network is boosted by using key frames that were extracted via applying Kalman filter during dataset augmentation. The proposed system resulted in promising performance compared to the state of the art approaches. The classification accuracy reached 92.35%.
N. Mostafa, N., F. Alrahmawy, M., & Nomair, O. (2018). Video Analysis For Human Action Recognition Using Deep Convolutional Neural Networks. Mansoura Journal for Computer and Information Sciences, 14(1), 1-7. doi: 10.21608/mjcis.2018.311989
MLA
Nehal N. Mostafa; Mohammed F. Alrahmawy; Omaima Nomair. "Video Analysis For Human Action Recognition Using Deep Convolutional Neural Networks", Mansoura Journal for Computer and Information Sciences, 14, 1, 2018, 1-7. doi: 10.21608/mjcis.2018.311989
HARVARD
N. Mostafa, N., F. Alrahmawy, M., Nomair, O. (2018). 'Video Analysis For Human Action Recognition Using Deep Convolutional Neural Networks', Mansoura Journal for Computer and Information Sciences, 14(1), pp. 1-7. doi: 10.21608/mjcis.2018.311989
VANCOUVER
N. Mostafa, N., F. Alrahmawy, M., Nomair, O. Video Analysis For Human Action Recognition Using Deep Convolutional Neural Networks. Mansoura Journal for Computer and Information Sciences, 2018; 14(1): 1-7. doi: 10.21608/mjcis.2018.311989
)
View on SCiNiTO