Red Delicious Apple Classification Based on Acoustic Response Changes during Storage Using Discrete Wavelet Transform and Artificial Neural Networks

Nowadays, new systems that are be able to measure the characteristics of the quality of food products for nondestructive testing and can be installed on the grading lines are very important. Analysis of the acoustic response is a non-destructive method for measuring fruit quality characteristics. Therefore, the potential of acoustic impulse response for non-destructive classification of Red Delicious apple was examined. In this study, sound signals in both time and frequency domains were analyzed using wavelet transform. Signals are decomposed into three levels using Daubechies 4. Eight statistical features were selected: maximum, minimum, mean, standard deviation, energy, kurtosis, skewness and third moment. Apples are classified according to changes in the acoustic response during storage time using multilayer perceptron neural network algorithm. According to results, classification performance of artificial network with 4-1-2 topology in time domain is better than the other networks. The classification accuracy and harmonic mean of precision and sensitivity for this topology were 82.1% and 0.81, respectively.