Human Behavior Recognition based on Temporally Segmentation
Lili Liang, Xu Liu, Yang Liu · 2024
The explosive growth of video streams poses challenges for performing human behavior recognition with high precision and low computational cost. Most of the existing methods have common problems: complex models and insufficient data samples, and the lack of feature extraction ability. For this problem, this paper proposes a temporally segmented human behavior recognition model based on Mixup data augmentation and CBAM. The model obtains image sequences through time-segmented sampling. Then, the ResNet-50 network with CBAM was used to extract the spatial features in the image sequence in parallel, and the Mixup technology was used to enhance the complexity of the data in the process of model training, so as to improve the recognition performance of the model. The experimental results show that the time-segmented human behavior recognition method based on Mixup and CBAM has achieved good recognition effect in the UCF101 and HMDB51 datasets, reduces the overfitting of the network, and improves the recognition accuracy in the two data sets by 1.51% and 4.84% respectively compared with the state-of-the-art model.