Novelconvolution Based Signal Processing Techniques for a Simplified Artificial Olfactory Mucosa

As our understanding of the human olfactory system increases, so does our ability to design novel architectures in order to mimic the biological system. The concept of an artificial olfactory mucosa represents a new development in the field of biomimetics. Here we analyse the signals produced by such a biomimetic system that contain a spatio-temporal element not previously encountered within the field of machine olfaction or so-called electronic noses. This paper explores the use of convolution-based signal processing methodologies to exploit this richer data-set and ameliorate the well-known problems of sensor noise and drift. We show that, under certain conditions, an artificial mucosa combined with a convolution based classifier performs better than a conventional electronic nose.