Estimating biomechanical measures of the human body requires body segment inertial parameters (BSIP). Widely cited estimation methods (Chandler et al., 1975;​ Dempster, 1955) are based on an elderly population, have inherent errors (Hanavan, 1964) and may not represent subjects to which they are applied. Subject specific methods such as medical imaging or geometrical techniques can be time consuming, expensive and may require a variety of manually acquired inputs. Presented here is an alternative, subject specific approach to estimating BSIPs, requiring no manual measurements at a equipment cost of approximately $300.

A 3D scanning protocol using a single Microsoft Kinect V2 was developed. Twenty-one subjects (10 male, 11 female) were scanned 3 times with per scan durations of approximately 20-30 seconds. Each full body 3D scan was segmented in Meshlab according to a 16 segment link model defined by Zatsiorsky and Seluyanov (1983). The 3D data was analyzed in custom MATLAB software that estimated the volume, MOI, COM, length and mass of each of the link segments. Results were compared to estimates obtained using an elliptical cylinder model (ECM) (Jensen 1978) and to estimates obtained using regression equations developed from medical imaging techniques (Zatsiorsky and Seluyanov 1983).

The body volume estimates using the Kinect demonstrated high repeatability (ICC≥0.95) with a volume overestimation ranging from 0.0023 m³ (3%) for males and 0.0038m³ (5.1%) for females. The upper torso accounted for the majority of the overestimation of approximately 0.0019m³ (83%) for males and 0.002 m³ (53%) for the females. The Kinect protocol estimated COM with an average error of <2%. Segmental mass estimates were overestimated by 3.11% for males and 5.23% for females when compared to ECM. This research established that the Kinect V2 could be used as a tool for BSIP estimation. Future work should improve on the scanning posture, scanning protocol and reducing the scan duration. Furthermore, future research should investigate on methods to eliminate the manual segmentation proposed for this protocol.