Neural-network-integrated electronic nose system for identification of spoiled beef

A commercially available Cyranose-320 (TM) conducting polymer-based electronic nose system was used to analyse the headspace from fresh beef strip loins (M. Longissimus lumborum) stored at 4 degrees and 10 degrees C. The raw signals obtained from the electronic nose system were pre-processed by various signal-processing techniques to extract area-based features. Principal component analysis was subsequently performed on the processed signals to further reduce the dimensionalities. Classification models using radial basis function neural networks were developed using the extracted features. The performance of the developed models was validated using leave-l-out cross-validation method. The developed models classified meat samples stored at two storage temperatures into two groups, i.e., "unspoiled" (microbial counts < 6.0 log(10) cfu/g) and "spoiled" (microbial counts >= 6.0 log(10) cfu/g). Maximum total classification accuracies of 100% were obtained for both the samples stored at 10 and 4 degrees C. Classification models based on "Area scaled" feature showed higher accuracies than that obtained using "Area unsealed feature." (c) 2005 Swiss Society of Food Science and Technology. Published by Elsevier Ltd. All rights reserved.