Pain expression recognition based on pLSA model.

Area of Science:

• Computer Vision
• Medical Informatics
• Biomedical Engineering

Background:

• Accurate pain assessment is crucial for patient care but often relies on subjective measures.
• Automated pain expression recognition from video can provide objective and consistent evaluation.

Purpose of the Study:

• To develop and validate a novel automated system for recognizing and categorizing pain expressions from video sequences.
• To quantify pain into four distinct levels: no pain, slight pain, moderate pain, and severe pain.

Main Methods:

• Facial velocity information was extracted using optical flow techniques to characterize pain indicators.
• A bag-of-words model represented facial velocity features as visual words for pain expression.
• A probabilistic Latent Semantic Analysis (pLSA) model, enhanced with class label information, was employed for pain expression recognition.

Main Results:

• The proposed method demonstrated high effectiveness in pain expression recognition.
• Experimental results on a custom dataset showed an average recognition accuracy exceeding 92%.

Conclusions:

• The developed automated system accurately recognizes pain expressions from video sequences.
• The integration of facial velocity, bag-of-words, and a label-informed pLSA model offers a robust solution for objective pain assessment.