Recognition of filled walnuts and empty walnuts using acoustic signal processing

An intelligent walnut recognition system combining acoustic emissions analysis, decision tree and fuzzy inference system (FIS) was developed and tested. In data acquisition part, Fast Fourier Transform (FFT) of impact signals was measured. Feature was extracted in two ways: using time domain and FFT of impact signal. The 66% of samples were used for training and the remains were used for testing. In selection feature part, the most important feature selected was: average and the second frequency amplitude of FFT. The method is based on the feature generation by FFT and time domain, produce decision tree with J48 algorithm and classification by fuzzy rules. The output of J48 algorithm was employed to produce the crisp if-then rule and membership function (MF) sets. The structure of FIS classifier was then defined based on the crisp sets. The results showed that the total classification accuracy was 94.7%, and the proposed FFT-J48-FIS model can be used in separation of filled walnuts from empty walnuts.