id_1022. DO DYADS SYNCHRONIZE OR ARE WE JUST IMAGINING IT? NOVEL PIPELINE FOR DYADIC MOVEMENT ANALYSIS
Aleksandra Kołakowska1, Julia Słowicka1, Grzegorz Kaliński1, Aleksandra Hamny1, Jarosław Żygierewicz2, Agnieszka Pluta1,3
1 University of Warsaw, Faculty of Psychology, 2 Banacha St., Warsaw, Poland
2 University of Warsaw, Faculty of Physics, 5 Pasteura St., Warsaw, Poland
3 Institute of Physiology and Pathology of Hearing, Bioimaging Research Center, World Hearing Center, 10 Mochnacki St., Warsaw, Poland
INTRODUCTION: Biobehavioral synchrony in dyadic interaction refers to the coordinated, mutual, and reciprocal alignment of physiological states, behaviors, or emotional expressions. The study followed Self- and Co-Regulation (SECORE) protocol that captures the phases of the interaction at different levels of regulation and arousal.
AIM(S): To build a research pipeline that detects dyadic movement from video recordings of the experiment.
METHOD(S): The training dataset for the predictive model consists of the object detection data from the Common Object in Context (COCO) dataset and corresponding age labels. For each person in a video frame, the algorithm infers the position of 17 body keypoints and age group classification. The computer vision pipeline outputs additional measures, i.e., distance, velocity, keypoint speed, movement entropy, and similarity in movement direction (cosine similarity).
RESULTS: The algorithm achieved a high average classification accuracy when using five-fold stratified cross-validation on the test COCO dataset. We compared 30 caregiver and typically developing child dyads (TD-TD) with 20 caregiver child dyads in which the child had autism spectrum disorder (TD-ASD). Preliminary results suggest that children in the ASD group maintained distance from their caregiver more often than typically developing children. Interestingly, the mean distance in the TD-TD dyads was significantly greater than the distance in the TD-ASD dyads. What is more, TD-TD dyads had a higher frequency of interaction, with both the child and the caregiver approaching each other. Children in both groups moved more towards the caregiver with greater magnitude variation in TD-TD dyads.
CONCLUSIONS: Taken together, these preliminary findings point to group-specific differences in proximity management and interaction dynamics that warrant further investigation in larger samples. Our results showed that the measures obtained from the pipeline can be used to analyse the dyadic synchrony which has a real impact on the field.
FINANCIAL SUPPORT: The research was supported by the European Union under the Horizon-Widera Europe program (grant agreement No. 101159414)