Abstract
A non-parametric estimate of the Frequency Response Function (FRF) of a measured system provides a lot of useful insight about that system.
This poster presents a frequency domain implementation of Gaussian processes to obtain a smoothed estimate of the FRF while, simultaneously, suppressing the transient errors. The hyperparameters of the gaussian process (as for instance the optimal smoothness) are learned from the data via cross validation.
A comparison with existing techniques, including the Local Polynomial Method (LPM) and a time domain regularised impulse response estimation, will be provided.
This poster presents a frequency domain implementation of Gaussian processes to obtain a smoothed estimate of the FRF while, simultaneously, suppressing the transient errors. The hyperparameters of the gaussian process (as for instance the optimal smoothness) are learned from the data via cross validation.
A comparison with existing techniques, including the Local Polynomial Method (LPM) and a time domain regularised impulse response estimation, will be provided.
| Original language | English |
|---|---|
| Publication status | Published - 22 Sept 2013 |
| Event | ERNSI 2013, Nancy, France, September 22-25, 2013 - Nancy, France Duration: 22 Sept 2013 → 25 Sept 2013 |
Conference
| Conference | ERNSI 2013, Nancy, France, September 22-25, 2013 |
|---|---|
| Country/Territory | France |
| City | Nancy |
| Period | 22/09/13 → 25/09/13 |
Keywords
- Frequency Response Function (FRF)
- estimation