15538
Capturing Social Motor Coordination in Children with Autism: Comparing the Microsoft Kinect, Video Analysis and Wireless Motion Sensor Tracking
Motor deficits have long been associated with autism but the role they play in social and emotional functioning is poorly understood. In addition, social movement coordination has been shown to facilitate social connection and may be abnormal in children with autism (Fitzpatrick et al., 2013). Recent advances in technology have resulted in a number of low cost gaming systems for remotely tracking human motor behavior that could potentially improve our understanding of bodily coordination exhibited by children with autism. Software development kits that enable the development of recording software that meet the specific needs of researchers interested in obtaining wireless time-series recordings of human movement has also made it much cheaper and easier to collect data. The degree to which these systems can replace expensive motion tracking systems, however, is unknown.
Objectives:
Here we present a comparison of skeletal data recorded using the Microsoft Kinect to data obtained using video analysis algorithms and a Polhemus Latus wireless motion tracking system. By comparing data recordings of various motor coordination behaviors obtained from a study on social motor coordination in typically developing children and children with autism, we detail the effectiveness of each system for studying social motor behavior in these populations.
Methods:
50 typically developing children and 50 children with autism from 72 to 131 months of age were asked to coordinate with a male experimenter in six tasks and to play pat-a-cake. The movements of the participants and the experimenter were recorded using a Kinect and a Polhemus system. From the Kinect we obtained both skeletal movement data and video. The time-series data obtained were then compared using dynamical time-series analysis.
Results:
All three data capturing methods revealed differences in the coordination that occurred for both groups. Children with autism exhibited less movement coordination with the experimenter overall. The magnitude of this difference, however, was task by method dependent. The most robust data collection method was the Polhemus system, which was able to capture differences in the fine grain patterning of motor coordination. The video analysis method provided the best measure of fully body coordination. The Microsoft Kinect skeletal data provided an adequate measure of global and, in some cases, local coordination, but overall was less robust than the other two.
Conclusions:
All sources proved useful in capturing motor coordination data. The type of task and experimental set-up, however, played a significant role in the success of each. The Polhemus system provided the most robust fine grain time-series data, but was limited to tasks that employed sensored limbs. Both the video analysis and the Kinect skeletal data provided better measures of global coordination, but were both strongly affected by occlusion. In order to obtain measures of movement for two separate individuals using video analysis or Kinect skeletal data participants’ movements must not overlap. To obtain useful Kinect skeletal data the participants must also face the recording systems more or less directly. Measuring motor coordination using these techniques may provide a better window into the underlying coordination problems in autism.