A Preliminary Study on Personalized 6-DoF Scapular Motion Recognition Using a Single-Device, Single-Site Optical MMG Array

This study investigates high-accuracy recognition of six degrees of freedom scapular motion for immersive VR flight with minimal user burden. Substantial inter-individual variability was quantified by contrasting a cross-subject model (40.33% mean accuracy) with a person-specific model (99.33%), revealing a large performance gap. Building on this, this study shows that a minimal configuration - a single optical mechanomyography (MMG) array at a serratus anterior-targeted window - achieves 97.33% accuracy with person-specific evaluation (10 trials per action). Learning-curve analysis further indicated that only 4–5 trials per action were sufficient to reach approximately 95% accuracy, supporting a short and practical calibration protocol. Early fusion of inertial measurement unit (IMU) features provided a consistent incremental gain of +1.67% points, improving accuracy to 99.00%. Accuracy was the primary evaluation metric because classes were balanced. These findings support a practical guideline: adopt “single optical MMG + personalization” as the default, and add IMU features only when needed. This work reports an engineering proof of concept focused on sensor placement and recognition; subjective outcomes such as embodiment and user experience are out of scope. Beyond the motivating VR flight scenario, the same minimal, personalized pipeline may extend to shoulder–scapular control in rehabilitation and sports, though such applications require dedicated validation.