![](bg9.png)
complex CMU-MMAC dataset (69% vs. 12%). This
is because the hierarchical classifier was built to group
and handle similar classes separately with specialized
features. Therefore, the more confusing dataset yields
a higher improvement.
However, we do recognize that although we out-
perform the non-hierarchical baselines, the resulting
accuracies are still low compared to previous work in
(Fisher and Reddy, 2011). This is because, instead of
focusing on the maximization of total accuracy as in
previous work, we focus on generating quality macro-
classes and testing the performance impact of using
the respective specialized feature sets. In an effort
to minimize the computational cost of our resulting
algorithm, we use computationally inexpensive sta-
tistical features and k-NN classification on the sec-
ond level of the hierarchy. Our accuracy would most
likely be substantially improved at the cost of com-
putational resources by using the more complex fea-
tures and classification methods of previous work on
the second level of the hierarchy. Once the test sample
has been correctly classified into a macro-class at the
top level (which we achieve very high performance),
we note that any type of feature set or classifier can
be used by the subsequent classification nodes.
Overall, we contribute a new algorithm to improve
the performance of the k-NN classifier by building a
hierarchical classification model with specialized fea-
ture selection. Our results show significant improve-
ment over the baseline, with the possibility to improve
further by using more complex features or classifiers
on the bottom level of the hierarchy.
ACKNOWLEDGEMENTS
Data used in this paper was obtained from
kitchen.cs.cmu.edu and the data collection was
funded in part by the National Science Foundation un-
der Grant No. EEEC-0540865.
REFERENCES
Ermes, M., Parkk, J., Mantyjarvi, J., and Korhonen, I.
(2008). Detection of daily activities and sports with
wearable sensors in controlled and uncontrolled con-
ditions. IEEE Transactions on Information Technol-
ogy in Biomedicine, 12(1).
Fisher, R. and Reddy, P. (2011). Supervised multi-modal
action classification. Technical report, Carnegie Mel-
lon University.
Ganti, R., Srinivasan, S., and Gacic, A. (2010). Multisensor
fusion in smartphones for lifestyle monitoring. In Pro-
ceedings of 2010 International Conference on Body
Sensor Networks.
Iso, T. and Yamazaki, K. (2008). Gait analyzer based on
a cell phone with a single three-axis accelerometer.
In Proceedings of 6th ACM Conference on Embedded
Networked Sensor Systems.
Karantonis, D., Narayanan, M., Mathie, M., Lovell, N., and
Celler, B. (2006). Implementation of a real-time hu-
man movement classifier using a triaxial accelerome-
ter for ambulatory monitoring. IEEE Transactions on
Information Technology in Biomedicine, 10(1).
Kim, K.-J., Hassan, M. M., Na, S., and Huh, E.-N. (2009).
Dementia wandering detection and activity recogni-
tion algorithm using tri-axial accelerometer sensors.
In Proceedings of the 4th International Conference on
Ubiquitous Information Technologies & Applications.
Kohavi, R. and John, G. H. (1997). Wrappers for feature
subset selection. Artificial Intelligence, 97(1-2).
la Torre, F. D. and Hodgins, J. (2009). Guide to the
carnegie mellon university multimodal activity (cmu-
mmac) database. Technical Report CMU-RI-TR-08-
2, Carnegie Mellon University.
Law, M., Figueiredo, M., and Jain, A. (2004). Simultaneous
feature selection and clustering using mixture models.
IEEE Transactions on Pattern Analysis and Machine
Intelligence, 26(9).
Liu, H. and Yu, L. (2005). Toward integrating feature selec-
tion algorithms for classification and clustering. IEEE
Transactions on Knowledge and Data Engineering,
17(4).
Miluzzo, E., Lane, N., Fodor, K., Peterson, R., Lu, H., Mu-
solesi, M., Eisenman, S., Zheng, X., and Campbell,
A. (2008). Sensing meets mobile social networks: The
design, implementation and evaluation of the cenceme
application. In Proceedings of 6th ACM Conference
on Embedded Networked Sensor Systems.
Saponas, T. S., Lester, J., Froehlich, J., Fogarty, J., and Lan-
day, J. (2008). ilearn on the iphone: Real-time human
activity classification on commodity mobile phones.
Cse technical report, University of Washington.
Spriggs, E., la Torre Frade, F. D., and Hebert, M.
(2009). Temporal segmentation and activity classifi-
cation from first-person sensing. In Proceedings of
IEEE Workshop on Egocentric Vision at Conference
on Computer Vision and Pattern Recognition.
Talavera, L. (2005). An evaluation of filter and wrapper
methods for feature selection in categorical clustering.
In Proceedings of 6th International Symposium on In-
telligent Data Analysis.
Taralova, E. (2009). Cmu multi-modal activity dataset an-
notations. In http://www.cs.cmu.edu/ espriggs/cmu-
mmac/annotations/.
Wang, Y.-C. F. and Casasent, D. (2008). New sup-
port vector-based design method for binary hierarchi-
cal classifiers for multi-class classification problems.
Neural Networks, 21(2-3).
Zeng, H. and Cheung, Y.-M. (2009). A new feature selec-
tion method for gaussian mixture clustering. Pattern
Recognition, 42(2).
PECCS 2012 - International Conference on Pervasive and Embedded Computing and Communication Systems
114