"Diabetes Mellitus (Diabetes)" is expressed as an increase in the blood sugar value due to insufficient secretion of insulin hormone released from the pancreas gland. Clinical findings reveals that diabetes can damage many vital organs such as eye, kidney, heart and nervous system."Diabetic Retinopathy" is one of the most common diabetes-related eye diseases. It is the leading cause of loss of vision in humans due to changes in the vessels of retinal layer that perceive the light behind the eye. In this study, features of horizontal and vertical Video-Oculography (VOG) signals from non-proliferative and proliferative diabetic retinopathy patients, have been used to classify the disease. Feature extraction was done by discrete wavelet transform. In classification process, feed forward artificial neural networks were used. In the training process, performance analysis was performed for the different segments obtained by cross validation method of the data set and the optimum data set segment was determined. The highest classification performance of proposed model was obtained by dividing the data set 80% for training - 20% for test. Our results revealed that proposed artificial neural network model is able to detect diabetic retinopathy disease from Video-Oculography (VOG) signals successfully.