A Posture Training System Based on Therblig Analysis and YOLO Model-Taking Erhu Bowing as an Example

Area of Science:

• Music Technology
• Computer Science
• Performing Arts Education

Background:

• Computer-assisted learning offers potential for analyzing musical performance and identifying areas for improvement.
• Traditional erhu instruction can benefit from objective feedback mechanisms to enhance technique.
• Objective analysis of erhu bowing technique is crucial for skill development.

Purpose of the Study:

• To develop a computerized system for erhu bowing training utilizing a You Only Look Once (YOLO)-OD model.
• To provide erhu learners with independent practice tools that offer performance analysis and real-time feedback.
• To assist educators in identifying and addressing specific erhu bowing challenges faced by students.

Main Methods:

• Therblig analysis was employed to segment and analyze the erhu bowing process into critical steps.
• A You Only Look Once (YOLO)-OD model was developed for detecting and tracking key objects in erhu bowing movements.
• Scoring methodologies for bow level and bow straightness were created and integrated into the training system.

Main Results:

• The developed system successfully detects and tracks critical erhu bowing components using the YOLO-OD model.
• The system provides objective scores for bow level and bow straightness, enabling self-assessment for learners.
• Real-time feedback and alerts for incorrect motions and postures facilitate immediate technique correction.

Conclusions:

• The computerized erhu bowing training system enhances independent practice by providing detailed performance feedback.
• The system empowers erhu learners to refine their technique through objective scoring and real-time corrective guidance.
• Educators can leverage the system's data to offer targeted instruction and improve student learning outcomes.