The human-based multisensor fusion method for artificial nose and tongue sensor data

Presently, an increased interest is apparent for the development of integra ted human-like smell and taste sensing capabilities, e.g., for chemical, pa per pulp, food, and medicine applications. This paper will present an original sensor fusion method based on human exp ert opinions about smell and taste and measurement data from artificial nos e and taste sensors. The "electronic nose" consists of an array of gas sens ors with different selectivity patterns, signal handling, and a sensor sign al pattern recognition and decision strategy, The "electronic tongue." whic h was developed for the taste analysis of liquids is based on pulse voltamm etry. Measurement data from the artificial smell and taste sensors are used to produce sensor-specific opinions about these two human-like sensing mod alities. This is achieved by a team of artificial neural networks and conve ntional signal handling which approximates a Bayesian decision strategy for classifying the sensor information, Further, a fusion algorithm based on t he maximum likelihood principle provides a combination of the smell and, re spectively, taste opinions, into an overall integrated opinion similar to h uman beings. The proposed integrated smell- and taste-sensing method is then illustrated by an application of real world measurements in the food industry.