Abstract

Mountain papaya fruits (Vasconcella pubescens) were tested for firmness with a nondestructive acoustic method for 14 days after harvest. The response of each fruit was analyzed with the Fourier transform to obtain a firmness index (FI) based on the second resonant frequency and with the Short Time Fourier Transform (STFT) to obtain a spectrogram frequency centroid (FC) index. The indexes were processed with a support vector machine (SVM) learning procedure in which days since harvest was taken as the basic truth of ripeness which the measured indexes attempt to estimate. The analysis of the results demonstrate that different groupings of the days into classes to be estimated give widely varying recognition rates and that the best rates are obtained when the classes are delimited using prior knowledge. © 2006 International Federation for Information Processing.