Nondestructive apple quality assessment using acoustic-vibrational response and artificial neural network

In this study, acoustic and vibration response methods were used to non-destructively estimate the firmness of apples. Gala apples were stored at 0 and 20 ° C for 9 and 6 weeks and were tested for acoustic response every week during storage. During the test, the samples were placed on a special device designed and made for this test and stimulated with a gentle tap. Impact sound and vibration were received by the microphone and accelerometer and transmitted as analog signals to the computer sound card and then converted to digital signals. Digital signals were converted from time domain to frequency domain by Fourier transform in MATLAB software. The dominant frequencies of the sound and vibration signals were extracted and the firmness indices were obtained from special equations. The results of acoustic and vibration tests were compared with the results of puncture test. The correlation coefficient between puncture firmness and firmness index was more than 0.92, which was significant at the probability level of 1%. Also, the dominant acoustic and vibrational frequencies and mass of the samples were used as three features as single, binary and ternary using artificial neural network to estimate the shelf life of apples. The shelflife was estimated at one, two and three week intervals and the results of fusion of binary and ternary in different modes increased from 9% to 30% the accuracy of classification of individual features.