Paper in ICCV 2007 on “Structure from Statistics – Unsupervised Activity Analysis using Suffix Trees”

Citation

Abstract

Models of activity structure for unconstrained environments are generally not available a priori. Recent representational approaches to this end are limited by their computational complexity and ability to capture activity structure only up to some fixed temporal scale. In this work, we propose Suffix Trees as an activity representation to efficiently extract the structure of activities by analyzing their constituent event subsequences over multiple temporal scales. We empirically compare Suffix Trees with some of the previous approaches in terms of feature cardinality, discriminative prowess, noise sensitivity, and activity-class discovery. Finally, exploiting the properties of Suffix Trees, we present a novel perspective on anomalous subsequences of activities and propose an algorithm to detect them in linear time. We present comparative results over experimental data collected from a kitchen environment to demonstrate the competence of our proposed framework.

Links

• http://dx.doi.org/10.1109/ICCV.2007.4408894
• doi:10.1109/ICCV.2007.4408894

BibTeX (Download)

@inproceedings{2007-Hamid-SFSUAAUST,
title = {Structure from Statistics - Unsupervised Activity Analysis using Suffix Trees},
author = {R. Hamid and S. Maddi and A. Bobick and I. Essa},
url = {http://dx.doi.org/10.1109/ICCV.2007.4408894
},
doi = {10.1109/ICCV.2007.4408894},
year  = {2007},
date = {2007-10-14},
urldate = {2007-10-14},
booktitle = {IEEE International Conference on Computer Vision (ICCV)},
publisher = {IEEE Computer Society Press},
abstract = {Models of activity structure for unconstrained environments are generally not available a priori. Recent representational approaches to this end are limited by their computational complexity and ability to capture activity structure only up to some fixed temporal scale. In this work, we propose Suffix Trees as an activity representation to efficiently extract the structure of activities by analyzing their constituent event subsequences over multiple temporal scales. We empirically compare Suffix Trees with some of the previous approaches in terms of feature cardinality, discriminative prowess, noise sensitivity, and activity-class discovery. Finally, exploiting the properties of Suffix Trees, we present a novel perspective on anomalous subsequences of activities and propose an algorithm to detect them in linear time. We present comparative results over experimental data collected from a kitchen environment to demonstrate the competence of our proposed framework.
},
keywords = {activity discovery, activity recognition, computer vision, ICCV, IEEE},
pubstate = {published},
tppubtype = {inproceedings}
}