Technion-Behavioral-Data-Science-Symposium-2023

Poster titled Tracking the development of goal-directed movement, presented at the 1st Technion Behavioral Data Science Symposium (01/01/2023).

Eitan Hemed1,^, Tal Roth1 , Sagi Jaffe-Dax2, Hagit Hel-Or3 & Baruch Eitam1

^Corresponding Author - contact.

Affiliations: 1Department of Psychology, University of Haifa; 2School of Psychological sciences, Tel Aviv University; 3Department of Computer Science, University of Haifa


Example Output

A 3-month old participant, here MediaPipe is used as a benchmark pose estimation solution.


Get the poster here


Funding

This research is supported by a dataset collection grant from the Data Science Research Center (DSRC) at the University of Haifa.

Logo

Abstract

The project is a work in progress, which aims to characterize the development of planned movement during infancy by analyzing videos of infants at unstructured play. Despite extensive research over the last five decades, there still remain lacunas in the field. Most of the existing research on infant movement (specifically, directional reaching) was conducted using a controlled experimental paradigm rather than in ecological settings. Typically, a toy mobile was tied to the limbs of an infant using strings, and researchers recorded the change in activity in response to feedback generated by the mobile, when the strings were attached or detached. Other studies used marker-based estimations of movement and less structured measures (e.g., placing the infant in a toddler swing with a toy that is reachable only by a single hand). There is still a scarcity of studies using naturalistic settings, such as that proposed here. This is in stark contrast to the ubiquitous use of mobile phones for documenting infant children at play in their natural surroundings. The current work will utilize a large dataset of videos of infants at the ages of 2-6 months, freely playing with a toy-mobile in their parents’ homes, currently being collected. Given the dataset, we will use marker-less pose estimation methods to extract information about infant movements from the videos and identify age-related changes in movements. First, we will attempt to replicate previous findings relating to characteristics of movements at different ages (e.g., interlimb joint coordination), following which, we will explore the data for patterns which were not previously documented, but may occur in naturalistic settings when the infant is not constrained (e.g., limb independence; persistence in movement; different reaching based on end-goal as touching VS. hitting; etc.). The high ecological validity of the project is novel, enabling the potential to document general and replicable patterns, thus making it more impactful than a structured lab study. Finally, while the motivation for previous studies was primarily oriented towards clinical screening, the proposed project will tap the development and underlying mechanisms of the ability to perceive control over the physical environment (see below), and its relation to sensorimotor-predictable feedback (e.g., feedback from the toy mobile) being rewarding independently of any tangible outcomes. Understanding how we develop the ability to perceive our own control over the environment and act volitionally, can shed light on the maturation of the reward system.


To cite:

Hemed, E., Roth, T., Jaffe-Dax, S., Hel-Or, H., & Eitam, B., (2023). Tracking the development of goal-directed movement. In the 1st Behavioral Data Science Sympostion, Technion.