Inhalt des Dokuments
Conference, Symposium, and Workshop Papers
| Citation key | Ren2020SSP |
|---|---|
| Author | R. L. G. Cavalcante and S. Stanczak |
| Year | 2020 |
| Journal | IEEE Statistical Signal Processing Workshop 2020, Rio de Janeiro, Brazil, 12th-15th July, 2020 (to appear) |
| Abstract | Information about the angular power spectra (APS) of signals impinging on antenna arrays have important applications in massive MIMO systems such as user clustering, angle-of-arrival estimation, and channel covariance estimation in FDD systems, to cite a few. Current approaches for the estimation of APS can be divided into two main groups: model based methods and data driven methods. The former methods are able to produce reliable estimates with little side information and no training , but they do not exploit any information from datasets (if available) to improve the estimates. In contrast, pure data-driven methods can provided good performance without any knowledge about physical models, but they do not provide any guarantees of robustness against sudden changes in the propagation environment, a common occurrence in wireless systems. Against this background, we propose novel hybrid model and data driven algorithms that use both information about models and any available dataset. To this end, we modify projection and optimization methods for APS estimation by using an inner product (and hence the induced norm and metric) that is learned from data. The proposed algorithms are able to produce reliable estimates that exploit both statistical information available in datasets and model knowledge. As a result, they are able to provide robustness against relatively large changes in the propagation environment without the need to perform frequent training. |
Zusatzinformationen / Extras
Quick Access:
Schnellnavigation zur Seite über Nummerneingabe
Auxiliary Functions
This site uses Matomo for anonymized web analytics. More information and opt-out options under data protection.