Computational parallels between the biological olfactory pathway and its analogue 'the electronic nose': Part II. Sensor-based machine olfaction.

Over the last fifteen years, we have witnessed a rapid expansion in the development of artificial odour sensing systems, or so called 'electronic nose' systems. Whilst the power of this approach to flavour has undoubtedly been demonstrated by its recent application to various complex odours, it will be argued that the original research programme, aimed at developing a comparative model of the biological olfactory pathway, has degenerated into an attempt to obtain an ad hoc workable system, based around readily available sensor and pattern recognition (PARC) technologies. At the time, the first 'model' nose system reflected the limited understanding of sensory information processing carried out within the biological olfactory pathway. We are now presented with an opportunity to evaluate and re-assess the architecture for an electronic nose, in view of the recent advances in understanding the key processing principals exploited by the olfactory bulb and cortex in the identification and characterisation of molecular stimuli. In Part II of this paper, we examine the parallels that exist between the biological olfactory system and the electronic nose. It is shown that the two systems share many similarities in their architectures and other properties, such as odour delivery, nonspecific sensor/receptor response, sensor/receptor preprocessing and content addressable memory (CAM) function. Of particular importance, both systems need to overcome similar operating problems, such as sensor/receptor drift, degeneration and poisoning, limited sensor/receptor sensitivity, discrimination of odour quality invariant of intensity and also the identification of particular odour components within a mixture of background odours. Finally, a number of opportunities for improving the biological plausibility of electronic nose systems are suggested that may yield an improvement in performance.