<![CDATA[This research introduces a prototype drug delivery robot that is controlled entirely through finger gestures, without the need for direct touch or additional physical control devices. The system utilizes a Pi Camera connected to a Raspberry Pi to capture the image of the user's hand, which is then processed by MediaPipe Hands to perform detection and extraction of 21 point landmarks in real-time. The position and relationship between landmarks are converted into a five-bit binary vector that represents which finger is raised. This binary data is then sent via serial communication to the STM32 Nucleo microcontroller, which is tasked with translating the binary pattern into motion commands (forward, backward, turn right/left, stop) to drive the DC motor. The results of testing the gesture recognition system, which was performed 30 times for each command, showed a high success rate. The gesture Stop achieved perfect success (30/30), followed by 'Forward' and 'Right' with 10 successes, and 'Backward' and 'Left' with 9 successes. This test shows that the system is able to respond accurately to gesture commands at a distance of 20 to 250 cm. The robot was also able to execute all motion commands responsively and accurately according to the recognized gestures. This prototype proves that MediaPipe can be an efficient and reliable method to implement gesture-based robot control on resource-constrained embedded platforms, as well as potentially applied in healthcare environments to minimize physical contact.]]>
Copyrights © 2026
