Markerless tracking of head movement kinematics: testing a sensorimotor integration model
Assessing movement kinematics in naturalistic conditions is an important challenge in neuroscience. Here I studied head motion kinematics using DeepLabCut, a valuable tool for 3D markerless pose estimation that is based on deep neural networks. The input to this network are videos capturing the gait of several participants. The output consists of labeled videos in which each label represents the probability that a pixel belongs to a speci c body part. I used these videos to extract head motion kinematics. By means of these results, I studied the nding that predictability of head motion can be used to determine how the brain integrates sensory and motor signals. In particular, I found peaks in the motor-to-sensory noise graph that correspond approximately to toe-o and heel strike. These peaks indicate the moments during which head movement is not well predicted. Thus, sensory signals are expected to play a more crucial role during this time frame in the gait cycle.
Faculteit der Sociale Wetenschappen