Social Context Segmentation Analysis in a Primate Model of Autism Spectrum Disorder

Social interaction is a dynamic process in which multiple autonomous agents communicate each other via exchanges their internal state such as emotions and intentions. Based on recent advance in computational technology, without simplified hypothesis verification and condition control, it has become possible to approach the interactive dynamics itself.\n There are major progresses in the natural language analysis, such as extraction of morphemes, and constructive reproduction of free conversation. However, the computational understanding of nonverbal communication has not been revealed, but that is possibly a universal framework leading to humans and other social animals.\n In this research, we will conduct research on the following points: 1. Development of free moving body motion capture system We will develop marker less motion capture system based on dynamic estimation of skeleton model by using spatial scanning with depth camera, and obtain the detail time sequence of body part trajectory and angles. 2. Context segmentation analysis of behavioral sequence Applying unsupervised morphological analysis, we will extract meaningful segment units and their transitions from the behavior time series data.\n As a verification system, this measurement / analysis method is implemented in a model animal which is controlled local neural circuit activity. By computational describing the influence of neural dysfunction on social nonverbal communication in the model, we aim to contribute for understanding and support of neurodevelopmental disorders with social disability like autism spectrum disorders.新学術領域研究「人工知能と脳科学」第5回領域会