Abstract:

Eight kinds of food grade spices were detected by electronic nose and analyzed by a variety of pattern recognition
algorithms. The results showed that principal component analysis could entirely distinguish among 8 different kinds of spices.
Linear discriminant analysis could not successfully distinguish these spices with 3 of them overlapping though shortened
intra-class distances were reported. A 100% recognition rate was acquired, when the spices were detected by the model based
on discriminant factor analysis. The regression model built by partial least squares analysis was used to estimate unknown
concentration of infusions of 4 spices showing errors between 4.0% and 8.7%. Therefore, electronic nose combined with proper
recognition algorithms is able to differentiate among and identify spices and estimate their concentration.

Key words: electronic nose, pattern recognition, spices