Using the Microsoft Kinect to assess human bimanual coordination
Abstract
Optical marker-based systems are the gold-standard for capturing three-dimensional (3D) human kinematics. However, these systems have various drawbacks including time consuming marker placement, soft tissue movement artifact, and are prohibitively expensive and non-portable. The Microsoft Kinect is an inexpensive, portable, depth camera that can be used to capture 3D human movement kinematics. Numerous investigations have assessed the Kinect's ability to capture postural control and gait, but to date, no study has evaluated it's capabilities for measuring spatiotemporal coordination. In order to investigate human coordination and coordination stability with the Kinect, a well-studied bimanual coordination paradigm (Kelso, 1984, Kelso; Scholz, & Schöner, 1986) was adapted. Nineteen participants performed ten trials of coordinated hand movements in either in-phase or anti-phase patterns of coordination to the beat of a metronome which was incrementally sped up and slowed down. Continuous relative phase (CRP) and the standard deviation of CRP were used to assess coordination and coordination stability, respectively. Data from the Kinect were compared to a Vicon motion capture system using a mixed-model, repeated measures analysis of variance and intraclass correlation coefficients (2,1) (ICC(2,1)). Kinect significantly underestimated CRP for the the anti-phase coordination pattern (p <.0001) and overestimated the in-phase pattern (p<.0001). However, a high ICC value (r=.097) was found between the systems. For the standard deviation of CRP, the Kinect exhibited significantly higher variability than the Vicon (p < .0001) but was able to distinguish significant differences between patterns of coordination with anti-phase variability being higher than in-phase (p < .0001). Additionally, the Kinect was unable to accurately capture the structure of coordination stability for the anti-phase pattern. Finally, agreement was found between systems using the ICC (r=.37). In conclusion, the Kinect was unable to accurately capture mean CRP. However, the high ICC between the two systems is promising and the Kinect was able to distinguish between the coordination stability of in-phase and anti-phase coordination. However, the structure of variability as movement speed increased was dissimilar to the Vicon, particularly for the anti-phase pattern. Some aspects of coordination are nicely captured by the Kinect while others are not. Detecting differences between bimanual coordination patterns and the stability of those patterns can be achieved using the Kinect. However, researchers interested in the structure of coordination stability should exercise caution since poor agreement was found between systems.
Degree
M.S.
Advisors
Haddad, Purdue University.
Subject Area
Kinesiology|Biomechanics|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.