Sensor-Array Optimization Based on Mutual Information for Sanitation-Related Malodor Alerts

Abstract

There is an unmet need for a low-cost instrumented technology for detecting malodor around toilets and emerging sanitation technologies for onsite waste treatment. Our approach to an electronic nose for sanitation-related malodor is based on the use of electrochemical gas sensors, and machine learning techniques are utilized to optimize the sensor array and for odor classification. We screened 12 sensors for different vendors and target gases and recorded response to odorants from fecal specimen and from confounding good odors such as popcorn. The analysis by two feature selection methods based on mutual information indicates that the feature dimensionality can be reduced to five features extracted from only three sensors. A logistic regression classifier with five features achieved 74.8% accuracy and 84.2% F1 score in odor classification. These early results are promising, and they can potentially enable the optimized design of an integrated e-nose system for alerting malodor, and which can be utilized in public toilets and onsite waste treatment systems.