Online Scent Classification by Ion-Mobility Spectrometry Sequences

Abstract

For ion-mobility spectrometry (IMS)-based electronic noses (eNose) samples of scents are markedly time-dependent, with a transient phase and a highly volatile stable phase in certain conditions. At the same time, the samples depend on various environmental factors, such as temperature and humidity. This makes fast classification of scents challenging. The present aim was to develop and test an algorithm for online scent classification that mitigates these dependencies by using both baseline measurements and sequences of samples for classification. A classifier based on the K nearest neighbors approach was derived. The classifier is able to use measurements from both transient and stable phase, yields a label for the analyzed scent, and information on the trustworthiness of the returned label. In order to avoid the classifier being fooled by irrelevant features and to reduce the dimensionality of the feature space, principal component analysis was applied to the data. The classifier was tested with four food scents, each presented in two different ways to the IMS. By using baseline measurements, the misclassification rate was reduced from 20.0 to 13.3%. A second experiment showed that the used IMS type experiences device heterogeneity.

A comparison of online methods for change point detection in ion-mobility spectrometry data

“…This approach only fails if the majority of channels fails or when the algorithm detects change points on multiple channels that are far from the real change points. Analysis of the data set collected for this paper and the data set used in [14] showed that both situations are improbable. Therefore, in this paper the Matrix form CUSUM algorithm using averages of detected change points is used.…”

“…Several sources simply set the all initial parameters to 1 and the hazard rate to the prior belief on the frequency of change points, but no justification has been given in the literature. A quick study [25, p. 37] using the data sets [14] revealed that setting 𝜅 and 𝛼 to one, 𝜇 to sample mean and 𝛽 being equal to the hazard rate ensures that the algorithm always finds meaningful change points. Initializing all parameters to 1 results in failing to find any change point.…”

“…In the context of the present work, online change detection algorithms are needed for supporting classification algorithms. The IMS readings by an electronic nose (eNose) often display transient and stable phases [14]. A common approach with IMS data is to wait for the switch from transient phase to stable phase and to use only https://doi.org/10.1016/j.array.2022.100151 Received 5 November 2021; Received in revised form 8 March 2022; Accepted 1 April 2022 data from the stable phase for scent classification.…”

A Comparison of Various Algorithms for Classification of Food Scents Measured with an Ion Mobility Spectrometry

“…The classifiers were applied to two data sets collected from food samples. One collected from headspace, where the environmental conditions were controlled, and the other collected from an office desk without explicit attempts to control the environmental conditions (see, for example, [ 26 , 27 , 28 ] for detailed description of the data collection). The objective of the paper was to present and compare several analytical tools for improving the classification accuracy and speed for food scents also in the presence of environmental noise.…”