Computationally efficient environmental monitoring with electronic nose: A potential technology for ambient assisted living

Recently, ambient assisted living technologies have emerged to improve the quality of life of ageing populations. Identification of health-endangering indoor gases with a hardware-friendly solution may provide an early warning of unhealthy living conditions. Electronic nose technology, using an array of non-selective gas sensors, is a potential candidate to achieve this objective, but state-of-The-Art gas classifiers hinder the development of low-cost and compact solutions. In this paper, we introduce a very simple classifier that transforms the multi-gas identification problem into pair-wise binary classification problems. This classifier is based on the resultant sign of the difference between values of the sensors' features for all possible pairs of sensors in each binary classification problem. A classifier qualification metric is defined to evaluate its suitability with given data of the target gases. As a case study, experimental data of four health-endangering gases, namely, formaldehyde, carbon monoxide, nitrogen dioxide and sulfur dioxide, is acquired in the laboratory by developing an array of commercially available gas sensors fabricated by Figaro Inc. and FIS Inc. A classification accuracy of 94.56% is achieved in distinguishing the target gasses with our proposed classifier. This performance is comparable to that of computation intensive state-of-The-Art gas classifiers despite our classifier's simple implementation.