PhD student in Machine Learning-based Event Detection in Electromyography Signals (EMG for AR/VR/XR and muscle training/rehabilitation) in Zürich

The  is looking for a PhD student in Human-Computer Interaction at ETH Zurich. Our research will be focused on event detection and fine-grained activity detection using electromyography sensors as input. Supplemental input modalities will include inertial sensors (IMU). The research will focus on machine learning-based signal processing and time series analysis to detect moments of interest in EMG and IMU sensor data.

Based on the new detection methods, we will design interaction techniques for online scenarios. This can be in the areas of Augmented Reality and Virtual Reality to control an interactive interface as well as in the medical domain for rehabilitation, monitoring, and exercise training.

Following the high number of unrelated applications, here are some guidelines:

• We will only respond to applications for this role submitted through this platform. Applications through email will be ignored.
• Your motivation letter must relate to the specifics of this position, especially how your experience relates to EMG/EEG, signal processing, or similar.
• We won't be able to advance candidates to the interview if no experience with EMG/EOG/EEG is evident.

Electromyography (EMG) is a diagnostic technique used to assess the health of muscles and the nerve cells that control them. It involves recording and analyzing the electrical activity produced by skeletal muscles. EMG is recorded using sensors, typically surface electrodes placed on the skin over the muscles that detect the signals generated during muscle contraction. The data collected through EMG is essential for diagnosing conditions like muscular dystrophy, and neuromuscular disorders, and for monitoring muscle responses in various medical and research settings. In addition to its medical applications, EMG is increasingly utilized in human-computer interaction, particularly in developing sophisticated control systems for prosthetics, gaming, and interactive virtual environments. By interpreting the specific patterns of muscle activity, EMG provides a unique interface that can translate human intention into machine control.

We aim to enhance interaction capabilities in AR/VR and monitoring capabilities in interactive medical scenarios through the advanced analysis of electromyography (EMG) data. Therefore, the research in this PhD will focus on developing robust machine learning algorithms for signal processing and time series analysis. These algorithms are designed to detect specific events and detailed activities from the raw data provided by the multimodal sensors.

• Literature review of machine learning techniques on signal and time series analysis (specifically EMG), including Multivariate Time Series Analysis, Sequence Modeling, Dimensionality Reduction, Anomaly Detection, Temporal Pattern Recognition, Feature Extraction and Engineering, and Prediction Models.
• Method development for machine learning-based signal processing, including predictive coding, contrastive learning, augmentation approaches, and multimodal learning from auxiliary inputs.
• Review and further development of state-of-the-art efforts in active learning, transfer learning with user-specific finetuning, and online learning.
• Experiment design for validation methods, offline based on datasets and online for empirical validation.
• Present research findings at academic conferences and seminars, engaging with the wider scientific community for feedback and knowledge exchange.
• Collaborations with others in the  to integrate developed methods into broader project objectives, contributing to shared goals and interdisciplinary learning and applications that involve end-users and patients.
• Later on: Explore practical applications of the methods in areas like AR, VR, and medical technology, leading to the creation of functional prototypes or models.

ETH requirements:

• written and spoken fluency in English
• an excellent master's degree (MSc., M.Eng. or equivalent) in Computer Science, Electrical Engineering, Bioengineering, Robotics, or related

requirements for the position:

• strong interpersonal and communication skills
• experience in signal processing
• experience in machine learning for time series; highly beneficial: experience with online inference
• understanding of multimodal data processing
• understanding of EMG signals (or similar: EOG, ECG, EEG) and data coding
• optional: sensing/electrical engineering background to understand EMG signal acquisition
• optional but useful: experience with VR or AR, Unity/C#

Prior experience in conducting user evaluations is useful but not a must. Experience with interactive and real-time systems is also a plus.

We offer an exciting environment and to study in and work with. Beyond the lab, ETH Zürich has several internationally recognized research groups dedicated to interactive systems, Human-Computer Interaction, AR/VR, health, and machine learning. In our research, we often collaborate with other groups and departments as well as with several other institutions and companies in Switzerland and abroad.