Bachelor-Thesis presented by Daniel Alte from Frankfurt am Main, about the control of a robotic arm using OpenBCI.
1. Gutachten: Prof. Dr. Jan Peters
2. Gutachten: Dr. Guilherme Maeda
3. Gutachten: Rudolf Lioutikov
“Electromyography (EMG) measurements recently became accessible and inexpensive due to new hardware developed for general Brain Computer Interface (BCI) applications. Low-cost devices for EMG measurements, therefore, enable potential deployment of intelligent hand prosthetics for a wider range of people, for example, children in developing countries who cannot afford the required and constant replacement of actuated prosthetics. However, such inexpensive devices use passive surface electrodes and the limited processing power only allows for very basic signal processing. As a consequence, EMG signals are characterized by noise, drift, and low repeatability. This thesis investigates the use of machine learning methods, particularly classification methods, to validate the feasibility of a low-cost BCI device for controlling a robotic arm. Part of this study regards the experimental search for the location of the surface muscles of the forearm, that have to be found and measured for different movements. This thesis investigates preprocessing methods for the raw measurements with a bandpass filter and root mean square (RMS) over a time window. Finally, this thesis compares classification results with linear support vector machines, K nearest neighbors and a Gaussian classifier. In our findings, the support vector machine and the k nearest neighbor output the best results. Although low-cost electromyography devices measure less reliable signals, they are capable of classifying different hand movements.”
Link to paper: