This document summarizes research on recognizing electromyography (EMG) signals from hand gestures to control prosthetics using artificial neural networks. EMG signals were collected from muscles during two hand gestures. Thirteen features were extracted from the signals and used to train and test several neural networks with different training algorithms. It was found that networks using the Levenberg-Marquardt algorithm achieved the best performance, with over 90% classification accuracy and the fastest training times, making it most suitable for accurate and rapid prosthetic control based on EMG pattern recognition.