Implementación de un sistema de clasificación de gestos del brazo humano utilizando Myo Armband para mando a distancia de un brazo robótico de 3GDL

Abstract

This work presents a hand gesture recognition system of five gestures based on electromyographic signals EMG measured using Myo Armband. Gestures to recognize are the same default gestures as Myo Armband: wave in, wave out, fist, open, pinch. Every user must perform an acquisition routine prior using the system. This routine consists of acquiring the EMG signals for every gesture repeated 5 times. EMG signals are filtered and rectified using a low pass filter before classification. Dynamic Time Warping algorithm is used as a distance metric between EMG signals. System’s prediction is obtained from k nearest neighbors classifier. The teleoperation of a robotic arm of 3 DOF is also included. Gestures recognized by the classification system are used as activation signals. Every degree of freedom has as reference one of the Tait Bryan angles: yaw, pitch, roll. These angles are calculated based on the orientation signal measured by Myo Armband. A PID controller of angular position is implemented over every degree of freedom. PID controllers were tuned heuristically. The robotic arm was implemented using the Lego Mindstorms EV3.