Education

This thesis focuses on developing an automated acoustic measurement station for benchmarking and optimizing Active Noise Cancellation (ANC) on the OpenEarable platform. Using a GRAS 43AG artificial ear and an external sound source, the system will measure acoustic transfer functions under controlled conditions. A key challenge is synchronizing the OpenEarable microphones with the artificial ear's reference microphone to ensure accurate time- and phase-aligned measurements. Based on these measurements, algorithms for automatic feedforward ANC filter optimization will be developed, enabling reproducible performance evaluation and systematic filter tuning.

Can a tiny device worn in your ear recognize when you cough, sneeze, or clear your throat? Wearable devices are becoming increasingly capable of monitoring our health in everyday life. However, many health-related events are still tracked manually, for example through smartphone diaries. This is particularly challenging for allergy-related symptoms, which can occur unexpectedly and frequently throughout the day. In the ALERT research project at TECO (KIT), we explore how smart earables equipped with multiple sensors can be used to recognize such events automatically. Using data from openEarables 2.0, an ear-worn sensing platform, we investigate how machine-learning algorithms can distinguish symptoms such as coughing, sneezing, sniffing, and throat clearing from normal everyday activities. The goal of this thesis is to develop and evaluate machine-learning methods that can recognize these events from multimodal sensor data, contributing to the development of future unobtrusive health-monitoring systems.

HAR models often perform well on users represented during training but degrade substantially when applied to unseen users due to differences in individual movement patterns. However, it remains difficult to quantify how large this user-level domain shift is without relying on a trained model. This bachelor’s thesis investigates model-free statistical measures based on time- and frequency-domain sensor features, dimensionality reduction, and clustering/separability analysis. The goal is to develop an interpretable domain-shift metric that correlates with the performance drop of HAR models on unseen users and can help estimate generalization difficulty directly from the data.

HAR models often perform well on users represented during training but generalize poorly to unseen users due to differences in individual movement patterns. Collecting sufficient labeled data to retrain or personalize models for every new user is impractical. This master’s thesis investigates Diffusion Models (Flow Matching / Stochastic Interpolants) as a way to generate personalized sensor data from only limited user information. User prototypes will be used to condition and steer the generative process toward specific or previously unseen users. The resulting synthetic data will be used to improve cross-user generalization and enable few-shot personalization of HAR classifiers. Candidates should have strong deep learning foundations, Python and PyTorch skills, and an interest in generative modeling and domain adaptation.