Human action recognition based on spatial–temporal relational model and LSTM-CNN framework. (2022)

• Record Type: Journal Article
• Authors: Senthilkumar, N; Manimegalai, M; Karpakam, S; Ashokkumar, S.R; Premkumar, M
• Is Part Of: Materials today. Volume 57:Part 5(2022)
• Page Start: 2087
• Page End: 2091
• Publication Date: 2022
• Subjects: Action recognition -- Dilated bi-directional LSTM -- CNN; Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2021.12.004
• Languages: English
• ISSNs: 2214-7853

Abstract

Abstract: Due to the increasing popularity of human skeleton capture systems, many new methods for implementing skeleton-based action recognition has been proposed. Some of these include Long Term Memory and Convolutional Neural Networks. These methods can investigate the significant spatial-temporal information, but they are limited in their capacity to do so in real-world scenarios. In this paper, a new spatial-temporal model with a bi-temporal end-to-end framework is proposed. A novel structure is proposed to combine the functions LSTM and CNN. The structure uses the dependency model to build the skeleton data for the proposed network.