Discriminative metrics for gas classification with spike latency coding

A multi-sensor array of the gas sensors is used in order to improve the selectivity of a single sensor and obtain a unique signature. Typically, pattern recognition algorithms are used to find a relationship between the multi-sensor array response and odor class. Theses methods usually accompanied with high computational requirement. Recent results reveal that time of first spike coding exhibits fast and efficient odor identification with reduced computational cost. The objective of this paper is two fold. Firstly, we propose a new probabilistic discriminative metric for assigning an odor class to observed test pattern of first spikes of the sensors in the array. Secondly, we propose the decision boundary criteria for the spike distance algorithm that assesses the spike pattern by comparing its relative distance with training gases. The performance evaluation of these metrics is carried out through experimental data of three different gases. The results show that our proposed metrics display excellent performance as compared to existing pattern recognition algorithms.