Abstract:
|
The main existent tool to monitor chemical environ-
ments in a continuous mode is gas sensor arrays, which have been
popularized as electronic noses (enoses). To design and validate
these monitoring systems, it is necessary to make use of machine
learning techniques to deal with large amounts of heterogeneous
data and extract useful information from them. Therefore, enose
data present several challenges for each of the steps involved in
the design of a machine learning system. Some of the machine
learning tasks involved in this area of research include generation
of operational patterns, detection anomalies, or classification and
discrimination of events. In this work, we will review some of the
machine learning approaches adopted in the literature for enose
data analysis, and their application to three different tasks: single
gas classification under tightly-controlled operating conditions,
gas binary mixtures classification in a wind tunnel with two
independent gas sources, and human activity monitoring in a
NASA spacecraft cabin simulator. |