Design of a software tool to determine the impact of rheological and physical properties of foodstu on human sensory perception, using Arti cial Neural Networks
Tutor / director / evaluatorKUTTER, ALEX
Document typeMaster thesis (pre-Bologna period)
Rights accessRestricted access - confidentiality agreement
This thesis researches a new idea for predicting the impact of rheological and physical properties of foodstu on human sensory perception. The sensory attributes are estimated by means of an automatic learning procedure, based on Arti cial Neural Networks (ANN). Two data sets comprising rheological and physical measurements as well as the mouth feel attributes perceived by a sensory panel of yoghurt samples serve as data base for training and validation of the capability of ANN to predict sensory attributes with physical measurements. The set of variables is reduced by Stepwise Regression algorithm, using the resulting subset to nd the optimal con gurations of neural nets by means of a Resilient propagation algorithm. Oral viscosity, the perceived viscosity in the mouth, is appraised by means of ANN obtaining a set of nets which can estimate its value satisfactorily. The same data is evaluated by Psychophysical Models and Multi Linear Regression in order to compare the results of statistical methods with ANN ones. A software tool, which was called "hANNdy", is developed to automate the complete process of evaluation a set of data with ANN. All the simulations and practical results are included.
SubjectsRheology -- Mathematical models, Neural networks (Computer science), Food -- Sensory evaluation, Reologia -- Models matemàtics, Xarxes neuronals (Informàtica), Aliments -- Anàlisi sensorial
ProvenanceAquest document conté originàriament altre material i/o programari no inclòs en aquest lloc web
|PFC_Laia Balañà Ferré.pdf||Report||9.125Mb||Restricted access|