Abstract

A robotic arm capable of mimicking hand gestures through data analysis of the mechanical activity of the muscles, using wearable piezoelectric sensors was developed. Piezoelectric discs were placed in the key locations of forearm and gestures from the arm were predicted by the system using an approximator neural networks. Each approximator creates a characteristic curve and event data for each sensors correlating to each finger movement. Using a segmentator, the event can be recorded, then compared to the approximated data by calculating the root mean square error (RMSE) to find the most similar, identifying the gesture. After classification, the result is transmitted wirelessly using Bluetooth low energy (BLE) to a robotic hand to mimic the gesture.