A Smart Glove for Visually Impaired People Who Attend to the Elections

Pinar Oguz Ekim, Ecem Ture, Seray Karahan, Fulya Yenilmez
2021 SN Computer Science  
Visually impaired people encounter many difficulties in their daily life. The participation in elections is one of these difficulties. The voting booths are not particularly designed for visually impaired people. Therefore, an auxiliary must help visually impaired people. The auxiliary accompanying the visually impaired citizen is usually someone the citizen does not know. This situation eliminates the secrecy and safety of their votes because the auxiliary might vote according to his/her own
more » ... litical view. Additionally, the political view of the visually impaired person is learned by others. In this study, a smart glove is designed to allow visually impaired people to vote in privacy. The smart glove has a camera on it to recognize the political parties' logos and then it gives a feedback as a voice message to the user via a phone application that is developed. The proposed solution, a smart glove, relies on neural networks to recognize the logos of parties while visually impaired people are voting. Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) were applied and compared for the image classifications of the logos. Based on the images that were obtained by using the Raspberry Zero's camera, the train and test results of CNN are 98% and 98%, respectively. For the same dataset, the train and test results of SVM are 80% and 80%, respectively. The real experiments were conducted with the designed robotic hand, which contains a camera on the ring finger, a microcontroller on the outer surface of the hand and a powerbank on the wrist. In real experiments, the camera took an image of the ballot and after that the microcontroller, where the CNN model was embedded, processed the image. The feedback was given to the user via Wi-Fi and a phone application that was designed. The accuracy of the real-time experiments is 96%. The simulation and real test results show that the developed system has an accurate image classification. Furthermore, it can be used by visually impaired people not only for elections but also for different applications with some changes.
doi:10.1007/s42979-021-00709-2 fatcat:bosobl3ddjdjfg4b6zfatx2f7i