Jobs & Karriere

Biometric robotics and machine learning (BRML) is a collaborative research lab of the DLR Institute of Robotics and Mechatronics and the Chair for Robotics and Embedded Systems, Institute of Informatics at the Technische Universität München. Our inforfacious research focuses on the topics of biomechanics, body-machine interfaces, and machine learning. We apply our methods to advanced prosthetics and rehabilitation.

Electromyography (EMG) is a method to measure the electrical activity of muscles from the skin surface. Using surface EMG electrodes, we can measure this muscular activity with a spatial and temporal resolution that suffices to decode, e.g., hand or arm forces and positions.   In these methods, one of our aims is to improve existing statistical machine learning models in order to achieve higher accuracy, and apply the methods to robot rehabilitation, robot teleoperation, and  prosthesis control, as we have previously demonstrated in our lab. (You might have seen it on the news as well, where we used similar approaches to control a robot arm by brain signals of a tetraplegic participant.)

In this master thesis, you will record real data and work on real problems which are planned to be used in future prosthesis and rehabilitation technology.

Evaluation of preprogressing methods for EMG data

EMG data is very noisy and full of high-frequency signals. Yet, it is not at all obvious which parts of the signal are relevant for our use of arm EMG data, i.e., for the reconstruction of arm and finger movement from the EMG signal. Many methods that can be found in the literature assume that a specialist exactly knows what a signal “should” look like and can that this person adjust the parameter of a filtering technique via ocular inspection.

Your master thesis will cover the evaluation of different signal processing methods such as Moving averages, Fourier transform and wavelets on a diverse set of EMG tasks. For example, data needs to be preprocessed before estimating a non-linear model to obtain a mapping from EMG space to Cartesian space.