Sistem Pengenal Isyarat Tangan Untuk Mengendalikan Gerakan Robot Beroda menggunakan Convolutional Neural Network

Currently, Human and computer interaction is generally done using a remote control. This approach tends to be impractical for wheeled robot operation because it must always carry an intermediary tool during the operation. The application of hand gesture recognition using digital image processing techniques and machine learning in the control process of wheeled robots will facilitate the control of wheeled robots because control no longer requires an intermediary tool.In this study, hand image taken using a camera then will be processed using a single board computer to be recognized. The results of recognized are passed on to arduino leonardo and DC motor to control twelve wheeled robot movement. The method used in this study is contrast stretching for preprocessing and Convolutional Neural Network (CNN) for hand recognition. This method is tested with a variation of  bright 26-140 lux, the distance from the face to the camera is 120-200cm. Hand recognition systems using this method resulting accuracy 97,5%, precision 97,57%, sensitivity 97.5%, spesificity 99,77 and f1 score 97.45%.