E-Nose Pattern Recognition and Drift Compensation Methods

Background

An extensive amount of research has been conducted in the past to develop and improve pattern recognition systems and correct drift for e-noses. Pattern recognition and machine learning have been employed to analyse data generated by e-noses in order to classify and quantify sensor responses Al-Maskari, B´elisle, et al. (2016b); Al-Maskari, Guo, and Zhao (2016); Al-Maskari, Saini, and Omar (2010). Despite the extensive research, the development of a robust and reliable pattern recognition system still remains a research issue for the chemical sensor community. Such limitations are attributed to the low sensitivity and stability of gas sensors as well as the multi-dimensionality and complexity of the surrounding environments. Current pattern recognition systems for e-noses still suffer from drift effects and noise which greatly affect gas sensor selectivity and sensitivity. Because of drift effect, sensors will not give exactly the same response even when they are exposed to the same gas with the same concentrations. Therefore, a method must be considered to correct gas sensor drift.

This section will discuss different pattern recognition and drift correction techniques found in the literature. Figure 1 provides a taxonomy of pattern recognition and drift methodologies applied to drift correction.

Figure 1.

Taxonomy of pattern recognition and drift correction techniques applied to E-Nose