ISSN 2394-5125
 

Review Article 


ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao.

Abstract
Convolutional neural networks (CNN) have shown evidences of reliability on action recognition tasks. However, their ability has been limited to correctly decoding spatial data. Moreover, the action data exhibits a spatio temporal characteristics in real time. To this effect, recurrent neural networks (RNNs) had been successful in reconstructing temporal data from action sequences. Despite their strong reputation, RNNs fall short in decoding long sequences of data, which are part of real actions. Consequently, this problem is effectively solvable by Long Short-Term Memory (LSTM) networks. In this work, LSTMs are used to measure the relatively between action sequences that can impact positively in recognition. Specifically, we select geometric patterns drawn from skeletal time series data for experimentation on 3-layer LSTM network and achieved better results than CNNs and RNNs. We demonstrate this on skeletal data by using four challenging benchmark skeletal action datasets.

Key words: Human action recognition, Long short-term memory, Geometric features, Microsoft Kinect.


 
ARTICLE TOOLS
Abstract
PDF Fulltext
How to cite this articleHow to cite this article
Citation Tools
Related Records
 Articles by K. Vijaya Prasad
Articles by P.V.V. Kishore
Articles by O. Srinivasa Rao
on Google
on Google Scholar


How to Cite this Article
Pubmed Style

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao. ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK. JCR. 2019; 6(5): 279-285. doi:10.31838/jcr.06.05.44


Web Style

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao. ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK. http://www.jcreview.com/?mno=93282 [Access: August 17, 2021]. doi:10.31838/jcr.06.05.44


AMA (American Medical Association) Style

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao. ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK. JCR. 2019; 6(5): 279-285. doi:10.31838/jcr.06.05.44



Vancouver/ICMJE Style

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao. ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK. JCR. (2019), [cited August 17, 2021]; 6(5): 279-285. doi:10.31838/jcr.06.05.44



Harvard Style

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao (2019) ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK. JCR, 6 (5), 279-285. doi:10.31838/jcr.06.05.44



Turabian Style

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao. 2019. ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK. Journal of Critical Reviews, 6 (5), 279-285. doi:10.31838/jcr.06.05.44



Chicago Style

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao. "ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK." Journal of Critical Reviews 6 (2019), 279-285. doi:10.31838/jcr.06.05.44



MLA (The Modern Language Association) Style

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao. "ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK." Journal of Critical Reviews 6.5 (2019), 279-285. Print. doi:10.31838/jcr.06.05.44



APA (American Psychological Association) Style

K. Vijaya Prasad, P.V.V. Kishore,O. Srinivasa Rao (2019) ACTION RECOGNITION FROM TIME SERIES GEOMETRIC JOINT FEATURES ON LSTM NETWORK. Journal of Critical Reviews, 6 (5), 279-285. doi:10.31838/jcr.06.05.44