Compressed-Domain Based Camera Motion Estimation for Realtime Action Recognition

Chen, Huafeng; Li, Hongyang; Xu, Zengmin; Hu, Ruimin

doi:10.1007/978-3-319-24075-6_9

Cited by 4 publications

(3 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Kantorov and Laptev 7 improved the DT by making the optical flow as substitution the motion vectors. MF 7 works well when there are only Intra coded frames (I‐frames) and Predicted frames (P‐frames) in the video compressed bitstream, as the motion vectors in P‐frame are forward predicted information and roughly reflect the changes of object movement 24 . Once the video compressed bitstream abound with B‐frames, the action recognition rate of MF decreases sharply (as shown in Section 4.3).…”

Section: Related Workmentioning

confidence: 99%

Real‐time action feature extraction via fast PCA‐Flow

Chen

et al. 2019

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

Summary Action recognition is a research hotspot in the field of Internet of Things (IoT). Currently, local pixel‐domain spatiotemporal feature extraction methods have reached the state‐of‐the‐art action recognition performance on many challenging datasets. However, the poor computational complexity of these approaches prevents them from scaling up to real‐time applications. For solving this problem, we present a novel real‐time video feature extraction technique by exploiting the fast PCA‐Flow algorithm. Firstly, we down‐sample video images in form of grid. Based on the down‐sampling images, PCA‐Flow algorithm is used to calculate optical flow among adjacent images. The PCA‐Flow matrices are then expanded to the original video image size by using efficient gCLSR super‐resolution method to keep the inherent geometric structure of the optical flow. Finally, we compute action descriptors based on original pixel frames and the enlarged PCA‐Flow images. The proposed approach is validated on three challenging datasets: UCF50, Hollywood2, and HMDB51. Experimental results indicate that the proposed method is more efficient in computation and can achieve competitive quality than the state‐of‐the‐art methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Real‐time action feature extraction via fast PCA‐Flow

Chen

et al. 2019

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

show abstract

“…As analyzed above, we noticed confusion between the static and dynamic postures of frames. Therefore, the proposed system integrates the LSTM block to process the entire data that solves the above problem based on [6], [22]. The goal of the system is to recognize simple and common postures.…”

Section: A Overview Of Systemmentioning

confidence: 99%

Proposing Posture Recognition System Combining MobilenetV2 and LSTM for Medical Surveillance

2022

View full text Add to dashboard Cite

This paper proposes a posture recognition system that can be applied for medical surveillance. The proposed method estimates human posture using mobilenetV2 and long short-term memory (LSTM) to extract the important features of an image. The output of the system was a fully estimated skeleton. We used seven human indoor postures, including lying, sitting, crouching, standing, walking, fighting, and falling, and classified them. The output results are the extraction of the human skeleton and the corresponding labels for the poses. We first experiment with classification using machine learning. The system only achieves approximately 88% accuracy because it is not able to classify similar postures, such as standing and walking. This difference can be caused by the extraction of features for static images, and the machine learning classification algorithm has not reached accuracy with training data. Therefore, we proposed the integration of the LSTM model into the proposed system. LSTM learns the features of the skeleton and provides classification results for postures. As a result, our system improved the accuracy by up to 99%. Similar postures, such as standing and walking, have improved accuracy by up to 7%. In addition, we performed the system on the Jetson Nano hardware. The results show that it can run on a low-profile (44% CPU and 2.1 frames per second) that is capable of applications for remote patient monitoring devices.INDEX TERMS Openpose, long short-term memory, posture detection, skeleton model, intelligent healthcare.

show abstract

“…Local feature description of underlying action features is still a hotspot in human behavior recognition research in recent years. Researchers considered the changes in the motion field between frames and proposed various local spatiotemporal feature descriptions, such as STIP [8], MoSIFT [9,10], and dense trajectories [2,11].…”

Section: Manualmentioning

confidence: 99%

Deep ChaosNet for Action Recognition in Videos

et al. 2021

Self Cite

View full text Add to dashboard Cite

Current methods of chaos-based action recognition in videos are limited to the artificial feature causing the low recognition accuracy. In this paper, we improve ChaosNet to the deep neural network and apply it to action recognition. First, we extend ChaosNet to deep ChaosNet for extracting action features. Then, we send the features to the low-level LSTM encoder and high-level LSTM encoder for obtaining low-level coding output and high-level coding results, respectively. The agent is a behavior recognizer for producing recognition results. The manager is a hidden layer, responsible for giving behavioral segmentation targets at the high level. Our experiments are executed on two standard action datasets: UCF101 and HMDB51. The experimental results show that the proposed algorithm outperforms the state of the art.

show abstract

Compressed-Domain Based Camera Motion Estimation for Realtime Action Recognition

Cited by 4 publications

References 16 publications

Real‐time action feature extraction via fast PCA‐Flow

Real‐time action feature extraction via fast PCA‐Flow

Proposing Posture Recognition System Combining MobilenetV2 and LSTM for Medical Surveillance

Deep ChaosNet for Action Recognition in Videos

Contact Info

Product

Resources

About