Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SitPose: Real-Time Detection of Sitting Posture and Sedentary Behavior Using Ensemble Learning With Depth Sensor

Hang Jin, Xin He, Lingyun Wang, Yujun Zhu, Weiwei Jiang, Xiaobo Zhou

TL;DR

SitPose tackles sedentary health risks by real-time detection of sitting posture using Azure Kinect depth sensing and bone-joint features. It builds a 33,409-sample dataset from 36 participants, derives joint-angle features, and uses a soft-voting ensemble of SVM, DT, and MLP (and comparisons with GBDT and TabNet) to achieve an F1 of 98.1% for posture classification. The system is deployed for real-time feedback and sedentary reminders, with latency under 100 ms, and demonstrates practical potential for ergonomic interventions. Limitations include sample size and controlled environments, with future work aimed at expanding participant diversity, multi-angle recognition, and real-world workplace testing.

Abstract

Poor sitting posture can lead to various work-related musculoskeletal disorders (WMSDs). Office employees spend approximately 81.8% of their working time seated, and sedentary behavior can result in chronic diseases such as cervical spondylosis and cardiovascular diseases. To address these health concerns, we present SitPose, a sitting posture and sedentary detection system utilizing the latest Kinect depth camera. The system tracks 3D coordinates of bone joint points in real-time and calculates the angle values of related joints. We established a dataset containing six different sitting postures and one standing posture, totaling 33,409 data points, by recruiting 36 participants. We applied several state-of-the-art machine learning algorithms to the dataset and compared their performance in recognizing the sitting poses. Our results show that the ensemble learning model based on the soft voting mechanism achieves the highest F1 score of 98.1%. Finally, we deployed the SitPose system based on this ensemble model to encourage better sitting posture and to reduce sedentary habits.

SitPose: Real-Time Detection of Sitting Posture and Sedentary Behavior Using Ensemble Learning With Depth Sensor

TL;DR

SitPose tackles sedentary health risks by real-time detection of sitting posture using Azure Kinect depth sensing and bone-joint features. It builds a 33,409-sample dataset from 36 participants, derives joint-angle features, and uses a soft-voting ensemble of SVM, DT, and MLP (and comparisons with GBDT and TabNet) to achieve an F1 of 98.1% for posture classification. The system is deployed for real-time feedback and sedentary reminders, with latency under 100 ms, and demonstrates practical potential for ergonomic interventions. Limitations include sample size and controlled environments, with future work aimed at expanding participant diversity, multi-angle recognition, and real-world workplace testing.

Abstract

Poor sitting posture can lead to various work-related musculoskeletal disorders (WMSDs). Office employees spend approximately 81.8% of their working time seated, and sedentary behavior can result in chronic diseases such as cervical spondylosis and cardiovascular diseases. To address these health concerns, we present SitPose, a sitting posture and sedentary detection system utilizing the latest Kinect depth camera. The system tracks 3D coordinates of bone joint points in real-time and calculates the angle values of related joints. We established a dataset containing six different sitting postures and one standing posture, totaling 33,409 data points, by recruiting 36 participants. We applied several state-of-the-art machine learning algorithms to the dataset and compared their performance in recognizing the sitting poses. Our results show that the ensemble learning model based on the soft voting mechanism achieves the highest F1 score of 98.1%. Finally, we deployed the SitPose system based on this ensemble model to encourage better sitting posture and to reduce sedentary habits.

Paper Structure

This paper contains 18 sections, 11 equations, 12 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Preliminary Knowledge
  3. Time of Flight
  4. Human Skeleton Joints
  5. Voting Classifier
  6. System Design
  7. System Overview
  8. Collection of Joint Coordinate Data
  9. Joint Angle Feature Extraction
  10. Composition of the Dataset
  11. Parameter Settings of Classification Models
  12. System Performance Analysis and Demonstration
  13. Model performance evaluation indicators
  14. Model Performance Comparison
  15. Model Deployment and Real-time Sitting Posture Recognition
  16. ...and 3 more sections

Figures (12)

  • Figure 1: Azure Kinect's spatial coordinate axis.
  • Figure 2: The framework of the sitting posture and sedentary habits detection system.
  • Figure 3: The illustration of six types of sitting posture and standing which we examine in this work.
  • Figure 4: Experimental environments in our study.
  • Figure 5: The recognized joints by the depth camera where the selected joints are marked in red.
  • ...and 7 more figures