Discovery of activity composites using topic models: An analysis of unsupervised methods

In this work we investigate unsupervised activity discovery approaches using three topic modelź(TM) approaches, based on Latent Dirichlet Allocationź(LDA), n -gram TMź(NTM), and correlated TMź(CTM). While LDA structures activity primitives, NTM adds primitive sequence information, and CTM exploits co-occurring topics. We use an activity composite/primitive abstraction and analyze three public datasets with different properties that affect the discovery, including primitive rate, activity composite specificity, primitive sequence similarity, and composite-instance ratio. We compare the activity composite discovery performance among the TM approaches and against a baseline using k -means clustering. We provide guidelines for method and optimal TM parameter selection, depending on data properties and activity primitive noise. Results indicate that TMs can outperform k -means clustering up to 17%, when composite specificity is low. LDA-based TMs showed higher robustness against noise compared to other TMs and k -means.

[1]  Sargur N. Srihari,et al.  Decision Combination in Multiple Classifier Systems , 1994, IEEE Trans. Pattern Anal. Mach. Intell..

[2]  B. Lo,et al.  Pattern mining for routine behaviour discovery in pervasive healthcare environments , 2008, 2008 International Conference on Information Technology and Applications in Biomedicine.

[3]  C. Elkan,et al.  Topic Models , 2008 .

[4]  E. Pitman.  SIGNIFICANCE TESTS WHICH MAY BE APPLIED TO SAMPLES FROM ANY POPULATIONS III. THE ANALYSIS OF VARIANCE TEST , 1938 .

[5]  John D. Lafferty,et al.  Correlated Topic Models , 2005, NIPS.

[6]  Stefano Soatto,et al.  Detecting Humans via Their Pose , 2006, NIPS.

[7]  W. Eric L. Grimson,et al.  Unsupervised Activity Perception by Hierarchical Bayesian Models , 2007, 2007 IEEE Conference on Computer Vision and Pattern Recognition.

[8]  Li-min Xia,et al.  The Complex Action Recognition via the Correlated Topic Model , 2014, TheScientificWorldJournal.

[9]  Irfan A. Essa,et al.  Unsupervised Activity Discovery and Characterization From Event-Streams , 2005, UAI.

[10]  Lawrence B. Holder,et al.  Discovering Activities to Recognize and Track in a Smart Environment , 2011, IEEE Transactions on Knowledge and Data Engineering.

[11]  Daniel Gatica-Perez,et al.  Discovering human routines from cell phone data with topic models , 2008, 2008 12th IEEE International Symposium on Wearable Computers.

[12]  Bernt Schiele,et al.  Discovery of activity patterns using topic models , 2008 .

[13]  Yang Wang,et al.  Human Action Recognition by Semilatent Topic Models , 2009, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[14]  Michael I. Jordan,et al.  Latent Dirichlet Allocation , 2001, J. Mach. Learn. Res..

[15]  Katayoun Farrahi,et al.  Extracting Mobile Behavioral Patterns with the Distant N-Gram Topic Model , 2012, 2012 16th International Symposium on Wearable Computers.

[16]  Irfan A. Essa,et al.  A novel sequence representation for unsupervised analysis of human activities , 2009, Artif. Intell..

[17]  Diane J. Cook,et al.  Discovering frequent user--environment interactions in intelligent environments , 2011, Personal and Ubiquitous Computing.

[18]  Donald E. Brown,et al.  Health-status monitoring through analysis of behavioral patterns , 2005, IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans.

[19]  Gerhard Tröster,et al.  Assessing Topic Models: How to Obtain Robustness? , 2012 .

[20]  John C. Tang,et al.  Rhythm modeling, visualizations and applications , 2003, UIST '03.

[21]  Irfan A. Essa,et al.  Augmenting Bag-of-Words: Data-Driven Discovery of Temporal and Structural Information for Activity Recognition , 2013, 2013 IEEE Conference on Computer Vision and Pattern Recognition.

[22]  Alex Pentland,et al.  Unsupervised clustering of ambulatory audio and video , 1999, 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99 (Cat. No.99CH36258).

[23]  Paul Lukowicz,et al.  Collecting complex activity datasets in highly rich networked sensor environments , 2010, 2010 Seventh International Conference on Networked Sensing Systems (INSS).

[24]  Gerhard Tröster,et al.  Recognition of User Activity Sequences Using Distributed Event Detection , 2007, EuroSSC.

[25]  David M. Blei,et al.  Supervised Topic Models , 2007, NIPS.

[26]  Daniel Gatica-Perez,et al.  Discovering routines from large-scale human locations using probabilistic topic models , 2011, TIST.

[27]  Juan Carlos Niebles,et al.  Unsupervised Learning of Human Action Categories , 2006 .

[28]  John D. Lafferty,et al.  Visualizing Topics with Multi-Word Expressions , 2009, 0907.1013.

[29]  Jian Lu,et al.  An unsupervised approach to activity recognition and segmentation based on object-use fingerprints , 2010, Data Knowl. Eng..

[30]  David B. Dunson,et al.  Probabilistic topic models , 2011, KDD '11 Tutorials.