A Scene Recognition and Semantic Analysis Approach to Unhealthy Sitting Posture Detection during Screen-Reading.

Area of Science:

• Robotics and Human-Computer Interaction
• Computer Vision and Machine Learning
• Biomedical Engineering and Health Informatics

Background:

• Prolonged unhealthy sitting postures pose significant health risks, including lumbar and cervical diseases and myopia.
• Current posture recognition methods in robotic systems often fail in complex environments due to a lack of scene context understanding.
• Automatic vision-based detection of unhealthy sitting postures is crucial for proactive health interventions.

Purpose of the Study:

• To develop an advanced method for detecting unhealthy sitting postures during screen-reading by integrating scene recognition and semantic analysis.
• To overcome the limitations of existing methods that focus solely on human skeletal features.
• To enhance the accuracy and robustness of posture detection in complex, real-world environments.

Main Methods:

• Utilized a Microsoft Kinect sensor for detecting and tracking key skeletal points of the human body.
• Employed the Faster R-CNN deep learning model for accurate object detection and feature extraction within the scene.
• Integrated semantic analysis via Gaussian-Mixture behavioral clustering for comprehensive scene understanding.
• Fused skeletal and scene-based features for robust discrimination of various sitting postures.

Main Results:

• The proposed method demonstrated high accuracy and effectiveness in detecting diverse unhealthy sitting postures during screen-reading.
• Successfully avoided erroneous detection in complicated environmental settings.
• Outperformed existing methods by identifying a broader range of unhealthy postures, including those previously undetectable.

Conclusions:

• The developed scene recognition and semantic analysis approach significantly improves the accuracy and scope of unhealthy sitting posture detection.
• This technology holds potential for integration into healthcare robotic systems as a medical assistance tool.
• The findings contribute to advancing behavior analysis in robotic systems for improved health monitoring and intervention.