One of the efficient, cost-effective, and rapid methods for non-destructive quality evaluation of horticultural and agricultural products is the acoustic analysis response generated by impact. In this research, the possibility of using sound processing resulting from the collision of hazelnuts with a galvanized iron plate for separating large, medium, and small hazelnuts, as well as separating nutty and hollow types, has been investigated. First, hazelnuts were divided into four groups based on size: large, medium, small, and empty. The hazelnuts of each group were weighed to separate the nutty and hollow types from the composition. Based on this approach, a system was designed and developed using a single-impact method to generate sound from hazelnuts upon impact with an iron surface. Artificial intelligence techniques were employed to analyze and classify the product based on features extracted from their acoustic signals. Then, in order to take samples, all hazelnuts were dropped from a height of 35 and 45 cm and hit an iron plate. The sound of the collision was recorded by the microphone. The recorded data is taken from the time domain to the frequency domain by Fast Fourier Transform (FFT) and graded using the features obtained from the gradation. The results of audio tests show that by comparing the direct ratio of weight with amplitude and sound pressure level, a hazelnut that has a pass range of less than 0.121Pa, and a sound pressure level of less than 75.2 dB is hollow or has an immature kernels. The study also revealed that the drop height did not significantly affect the quality classification, and a height of 35 cm, along with a sound pressure level above 75.2 dB, yielded the best results for hazelnuts classification. The results of this research, in combination with the common methods of separating hazelnuts, show that acoustic response analysis as a non-destructive method of grading hazelnuts can reduce their damage in addition to increasing the speed and accuracy of calssification.