Navigation

Signal Processing

We use and develop a variety of signal processing tools with applications in physiological and movement assessment.

IMU sensor fusion and motion tracking

Inertial measurement units (IMUs), containing an accelerometer and a gyroscope are a cost-effective and unobtrusive wearable solution to measure human movement. However, raw sensor output data like acceleration and rotation rate provide only indirect and hard-to-interpret information. Therefore, we develop algorithms extract interpretable information about performed movements or even movement quality. This information can then be used to automatically classify movements, enable novel methods to interact with computer systems, calculate medical parameters, or directly monitor the health state of a patient. With these approaches we can convert IMU-based wearables, or even a modern smartphone, into a capable medical diagnosis tool, a sports tracker, or interactive gaming platform.

Analysis of long term monitoring data

In contrast to traditional lab data, which is recorded in short and very controlled settings under the supervision of a technical or clinical expert, the analysis of long-term monitoring data bares some additional challenges. The large amounts of data collected over multiple days or weeks ideally contain real life movements of participants or patients, meaning literally everything can be happening in the respective signals. As most often no reference gold standard data is available, smart algorithmic tools are required to obtain the useful information required for a scientific research question and reject the remaining data. Depending on the domain, we then apply algorithms to extract the required information from the data, for example stride to stride parameters in gait analysis or heart rate in ECG. For certain questions it might be useful to apply innovative strategies to re-evaluate and summarize the computed values. In case of gait analysis this can include an approach for assembling walking bouts on selected strides in order to provide summary statistics for several selected walking bouts over meaningful periods of time (hour, day, week, …).

Development of software tools for movement analysis

Movement analysis, in particular using wearable devices, has gained a massive interest in the research community. However, as in many rapidly growing fields, the technology available in consumer applications lags behind state-of-the-art research by a couple of years. Easy access to reliable implementation of popular algorithms is a crucial component to bring any research to the masses. Therefore, we are committed to develop our algorithms in a way that other researchers can easily use them. Whenever feasible, we also share our algorithms as open-source software. On a higher level, we are involved in and lead internal and EU-wide projects that aim to collect, reimplement, and standardise existing algorithms for medical movement analysis.

Gait analysis in animal models

By using signal processing approaches, the information of mouse paw positions can be used for analyzing mouse gait. This is especially important for analyzing gait progression in disease model animals. We develop novel approaches to analyse respective gait parameters, e.g. for the assessment of sway-related gait movement in Parkinson’s-Disease-relevant mice.

Cardiovascular monitoring

We acquire and process electrophysiological signals such as ECG and PPG to analyze cardiovascular activity (heart rate, heart rate variability, blood pressure). In particular, we are interested in the cardiovascular response to postural changes to detect autonomic failure, a possible indicator for autonomic disorders like Parkinson’s Disease or Multiple System Atrophy.

Analysis of acute and long-term stress markers

To better understand the link between acute as well as chronic stress and human health we record and process electrophysiological signals (ECG, EDA, EEG) as well as endocrinological markers (salivary cortisol or alpha-amylase). Among many, particular goals are to classify and predict different patterns of acute stress (some of which may be indicators of the development of chronic stress) and to explore methods to reduce acute stress.

 

Further information can be found in the projects of the following groups: