GAS-GCN: Gated Action-Specific Graph Convolutional Networks for Skeleton-Based Action Recognition.

Wensong Chan,Zhiqiang Tian,Yang Wu

doi:10.3390/s20123499

Abstract

Skeleton-based action recognition has achieved great advances with the development of graph convolutional networks (GCNs). Many existing GCNs-based models only use the fixed hand-crafted adjacency matrix to describe the connections between human body joints. This omits the important implicit connections between joints, which contain discriminative information for different actions. In this paper, we propose an action-specific graph convolutional module, which is able to extract the implicit connections and properly balance them for each action. In addition, to filter out the useless and redundant information in the temporal dimension, we propose a simple yet effective operation named gated temporal convolution. These two major novelties ensure the superiority of our proposed method, as demonstrated on three large-scale public datasets: NTU-RGB + D, Kinetics, and NTU-RGB + D 120, and also shown in the detailed ablation studies.

Highlights

With the wide use of the sensors like cameras and wearable devices, more and more information about our life is recorded
To the best of our knowledge, this is the first work to adopt gated Convolutional neural networks (CNNs) in temporal dimension for action recognition. Integrating these two major novelties, we propose a new model named gated action-specific graph convolutional network (GAS-graph convolutional networks (GCNs)), in which action-specific graph convolutional module (ASGCM) is applied to process the spatial information and the gated convolutional neural network operates in the time dimension
We propose an action-specific graph convolutional module, which combines the structural and implicit edges together and automatically learns the ratio of them according to the input skeleton data

Summary

Introduction

With the wide use of the sensors like cameras and wearable devices, more and more information about our life is recorded. These sensory signals contain plenty of human action information. How to analyse these data to recognise human action has become a popular issue in recent years. Skeleton-based action recognition has many applications in our life, such as human-computer interaction, intelligent monitoring, and industrial robots. Many methods for skeleton-based human action recognition are proposed in recent years [1,2,3,4,5,6,7]. Convolutional neural networks (CNNs) are abroad leveraged to extract the spatial-temporal features and generate classifications of action [8,9,10]

Methods

Results

Conclusion