- Project duration: 12 months (project start 2018-08-13)
- Partners: Department of Mathematics, Prof. Markus Haltmeier, University of Innsbruck
- Funding: € 10.000
Title of the project:
- Deconvolution algorithms for signal reconstruction
- In order to fully exploit the potential of our laboratory equipment, we need to correct our data by the sensor delay. The standard algorithms for this correction have the disadvantage that they amplify an existing sensor noise disproportionately strong. We therefore need a partner who has specialist knowledge in the field of signal processing and signal reconstruction. The Innovationscheck PLUS allows us to hire a proven expert in this field.
Contribution of our scientific partner:
- Our scientific partner will produce two example implementations of a deconvolution algorithm tailored to our problem: one with an iterative reconstruction method, and one with a deep-learning method based on neural folding networks. The iterative method is less complex to implement and better established, but modern deep learning methods are potentially numerically more efficient and even more accurate. Based on the sample implementations and their documentation, we will be able to integrate these algorithms into our software.
- The new deconvolution algorithm will help us not to unnecessarily degrade the signal-to-noise ratio and to eliminate systematic effects of sensor delay, mainly in oxygen kinetics experiments. As a result, our devices will become even more precise.