A SIGNAL REPRESENTATION APPROACH FOR DISCRIMINATION BETWEEN FULLAND EMPTY HAZELNUTS

摘要

We apply a sparse signal representation approach to impactacoustic signals to discriminate between empty and fullhazelnuts. The impact acoustic signals are recorded bydropping the hazelnut shells on a metal plate. The impactsignal is then approximated within a given error limit bychoosing codevectors from a special dictionary. Thisdictionary was generated from sub-dictionaries that areindividually generated for the impact signals correspondingto empty and full hazelnut. The number of codevectorsselected from each sub-dictionary and the approximationerror within initial codevectors are used as classificationfeatures and fed to a Linear Discriminant Analysis (LDA).We also compare this algorithm with a baseline approach.This baseline approach uses features which describe the timeand frequency characteristics of the given signal that werepreviously used for empty and full hazelnut separation.Classification accuracies of 98.3% and 96.8% were achievedby the proposed approach and base algorithm respectively.The results we obtained show that sparse signalrepresentation strategy can be used as an alternativeclassification method for undeveloped hazelnut separationwith higher accuracies.