Odor classification by a robot equipped with an
electronic nose (e-nose) is a challenging task for pattern
recognition since volatiles have to be classified quickly and
reliably even in the case of short measurement sequences,
gathered under operation in the field. Signals obtained in
these circumstances are characterized by a high-dimen-
sionality, which limits the use of classical classification
techniques based on unsupervised and semi-supervised
settings, and where predictive variables can be only iden-
tified using wrapper or post-processing techniques. In this
paper, we consider generative topographic mapping
through time (GTM-TT) as an unsupervised model for
time-series inspection, based on hidden Markov models
regularized by topographic constraints. We further extend
the model such that supervised classification and relevance
learning can be integrated, resulting in supervised GTM-
TT. Then, we evaluate the suitability of this new technique
for the odor classification problem in robotics applications.
The performance is compared with classical techniques as
nearest neighbor, as an absolute baseline, support vector
machine and a recent time-series kernel approach, dem-
onstrating the eligibility of our approach for high-dimen-
sional data. Additionally, we exploit the learning system
introduced in this work, providing a measure of the rele-
vance of each sensor and individual time points in the
classification process, from which important information
can be extracted.