CNN architecture for robotic arm control in a 3D virtual environment by means of by means of EMG signals
This paper presents the development of a 3D virtual environment to validate the effectiveness of a Convolutional Neural Network (CNN) in a virtual application, controlling the movements of a manipulator or robotic arm through commands recognized by the network. The architecture of the CNN network was designed to recognize five (5) gestures by means of electromyography signals (EMGs) captured by surface electrodes located on the forearm and processed by the Wavelet Packet Transform (WPT). In addition to this, the environment consists of a manipulator of 3 degrees of freedom with a final effector type clamp and three objects to move from one place to another. Finally, the network reaches a degree of accuracy of 97.17% and the tests that were performed reached an average accuracy of 98.95%.