Application of Voice Recognition Technology in Sorting of Soybean Varieties

Soybean sorting devices based on discrimination of physical properties of seeds don’t have high sorting accuracy, are heavy and have high volume, produce too many noises and consume a lot of energy. Also use of them led to seed defections and may cause seed damage. The use of impact sound processing is a new technique in sorting of agricultural products. In this paper we used acoustic signals of the seeds impacted onto a steel plate from 25 cm of elevation to separateSahar variety from Viliams. The digitized sound signals were processed in time domain to extract suitable features for variety separation. The extracted features were used as inputs to a multilayer neural network (MLP). The results indicated that the 80-3-2 network was the most suitable network for this separation. The test results showed that the proposed neural network was able to detect Sahar variety with 90% of accuracy and Viliams variety with 80%.

کلیدواژه ها:

soybean ، impact sound signal ، voice recognition ، artifitial neural network