Statistical Assessment of Yoga Posture Acquisition using a 3D Room Sensor

Abstract

Many innovative information technology applications use gestures as input. We are exploring gesture analysis for incorporation into exergames for personalized medical interventions using yoga as therapy (YT). PURPOSE: A data-driven machine learning solution for gesture detection was used to classify captured yoga poses with high accuracy. The research goal is to test whether a machine learning algorithm in a basic computer video exergame can assess yoga skill acquisition in targeted select populations as a means to promote healthy physical activity. METHODS: Convenience sample of 20 adult students, male and female of any race/ethnicity, were briefly instructed and shown poses to perform, while recorded by the Kinect attached to a PC. Three yoga sessions (pre-test, mid-way and a post-test) were captured during the regularly scheduled yoga class which met twice weekly for 75 minutes, over a 10-week period. RESULTS: We recorded 6 yoga instructors while performing a series of yoga postures, and recorded clips were tagged or labelled in all of the frames in the recordings that defined a yoga gesture by consensus of two yoga instructors. Default settings produced solutions with high True Positives (99.5%) and low False Positives (0.03%) for most yoga postures sampled. Depth stream and skeleton coordinates for the 20 participants were acquired and analyzed against the previous trained solution. Analysis of summary statistics was done for five yoga poses comparing initial, mid-session, and final session captures. Sensitivity showed consistent trends for Mountain, Forward Bend, and Upward Salute. For Mountain, Sensitivity went from 0.78 to 0.87, while the expert test clip scored 0.94. Informedness also showed similar consistent trends for those poses. Based on these results the higher sensitivity score predicts greater training and closer the postures were to the “gold standard”. CONCLUSIONS: Gesture analysis for yoga alignment training may be a useful tool for the development of home and clinical yoga therapy for hard to reach populations. The experimental exergame developed here provides a tool that scores the performance of yoga postures and provides improvement metrics. Our plans are to target special aging populations with YT, and study the potential effects of body mass and age on posture alignment and limb stretch.