Novel Subject-Dependent Human-Posture Recognition Approach Using Tensor Regression

Publication Date

1-1-2024

Document Type

Article

Publication Title

IEEE Sensors Journal

DOI

10.1109/JSEN.2024.3493893

Abstract

A novel dynamic human-posture recognition approach using tensor regression is proposed in this work. In our proposed approach, a new dynamic segmentation scheme based on hidden logistic regression is first undertaken to segment multi-dimensional skeletal-graph data. Within each segment of multi-dimensional data, a new feature tensor consists of high-dimensional skeletal-graph time-series involving multi-joint three-dimensional coordinates and their temporal differences. Regression models can thus be trained from these collected feature tensors with respect to each type of human postures of interest. Experiments using real-world Kinect data are conducted to evaluate the effectiveness of our proposed novel tensor-based human-posture recognition scheme. In comparison with two prevalent deep-learning models, namely graph convolutional network (GCN) and transformer, our proposed novel tensor-based human-posture recognition approach can achieve the highest recognition accuracy of 97%. Furthermore, we have evaluated the performance of our proposed new method using the open-source Kinect dataset, namely UTKinect dataset, for one-shot learning. Our proposed novel tensor-based human-posture recognition approach still significantly outperforms the aforementioned prevalent deep-learning models for one-shot learning.

Funding Number

2020GXNSFAA159146

Funding Sponsor

National Natural Science Foundation of China

Keywords

dynamic multi-dimensional time-series segmentation, human-posture recognition, Kinect data, skeletal graph, tensor regression

Department

Applied Data Science

Link to Full Text

Plum Print visual indicator of research metrics
PlumX Metrics
  • Usage
    • Abstract Views: 1
see details

Share

COinS