Document Type : Original Paper

Authors

• Ehsan Mahmoudi ¹
• Seid Mahdi Jafari ²
• Neda Rahmanian Jahromi ¹
• Mohammad Javad Akbarian Meymand ¹
• Mehdi Vatankhah ³

¹ MSc. Graduated Student, Gorgan University of Agricaltural Science & Natural Resources, Gorgan, Iran

² Assistant Professor, Department of Food science and Technology, Gorgan University of Agricaltural Science & Natural Resources, Gorgan, Iran

³ MSc. Graduated Student, Sabzevar Branch, Islamic Azad University, Sabzevar, Iran

Abstract

In this study, Clash Voice processing in the field of wavelet and Artificial neural network (ANN) has been evaluated in order to separate hazelnuts according to their size (large, small, medium), and also being filled or empty. Hazelnuts were divided into three size groups (large, medium and small) and each hazelnut of the groups was weighted in order to separate the filled ones from the empty ones. All hazelnuts were then released from 40 and 50 cm height; bopped with a metal plate and Clash Voice was recorded. Signals in the field of wavelet were processed after the Preprocessing. Obtained profiles were applied as input to the ANN. The results showed that Neural Networks with 3 and 9 neurons in the hidden layer could successfully (accuracy of 100%) break up the hazelnuts that were dropped from 40 and 50 cm heights. The filled and empty hazelnuts of large, medium and small groups were separated from each other with the accuracy of 100, 99.61 and 98% for the height of 40 cm and 100, 99.66, 97.5% for the height of 50 cm. This research in combination with the common methods could probably reduce damages; increase the precision and the speed of separation.

Keywords

• Acoustic
• Artificial Neural Network
• Discrete Wavelet transform
• Hazelnut
• Separation
• Sound