Research on Short-Term Smell-Based Room Identification with a Single-Type eNose

• The smell of seven rooms was measured with a single sensor type eNose • An LDA classifier was used to differentiate between room samples • The classifier was able to differentiate smelling and non-smelling rooms • Three rooms with characteristic smell were identified • The four non-smelling rooms could not be differentiated This paper presents a research on short-term room identification based on its smell measured with a small and lightweight electronic nose (eNose) in an educational building. The eNose used is based on 16 single-type chemo-resistive metal-oxide (MOX) material gas sensors which are configured with different measurement parameters to enhance its variability and operate together as an eNose. An untrained human operator selected three rooms with a differentiable smell and four neutral non-smelling rooms in an educational building to create referent training and validation datasets. The eNose was used to measure the smell of the rooms for three consecutive days. A linear discriminant analysis (LDA) of the training dataset showed different variance axis in the smell measurements registered in the rooms. Finally, the LDA classification of the validation dataset has shown that the three smelling rooms can be differentiated using eNose measurements while the other four neutral non-smelling rooms are undifferentiable using the presented method. These results agree with the natural sense of smell of an untrained human operator.