While person-tracking systems can capture very fine-grained, accurate data, the creation of art pieces and interactive experiences making use of captured data often benefits from being able to work with higher-level features. We propose a computational framework for interpreting person-tracking data and publishing the resulting information over a network for use by client applications, and emphasize the recognition of patterns of movement, both over time and instantaneously. Our system consists of four modules for tracking instantaneous features, short-time features, and using unsupervised and supervised machine learning techniques to extract features at higher levels of abstraction. Data used by the system is collected using OpenPTrack, an open-source library for person and object tracking geared towards accessibility to the arts and education communities.
© Copyright 2018 Proceedings of the 5th International Conference on Movement and Computing. Published by ACM. All rights reserved.