%0 Journal Article %J IEEE Transactions on Signal Processing %D 2017 %T Learning Power Spectrum Maps from Quantized Power Measurements %A Daniel Romero %A Seung-Jun Kim %A Georgios Giannakis %A R. López-Valcarce %K cognitive radio %K compass %K compressed sensing %K spectrum sensing %K winter %K wsn %X
Using power measurements collected by a network of low-cost sensors, power spectral density (PSD) maps are con-
structed to capture the distribution of RF power across space and frequency. Linearly compressed and quantized power measure-
ments enable wideband sensing at affordable implementation complexity using a small number of bits. Strengths of data- and model-
driven approaches are combined to develop estimators capable of incorporating multiple forms of spectral and propagation prior
information while fitting the rapid variations of shadow fading across space. To this end, novel nonparametric and semiparametric
formulations are investigated. It is shown that the desired PSD maps can be obtained using support vector machine-type solvers.
In addition to batch approaches, an online algorithm attuned to real-time operation is developed. Numerical tests assess the performance of the novel algorithms. 
%B IEEE Transactions on Signal Processing %V 65 %P 2547-2560 %8 05/2017 %G eng %N 10 %R 10.1109/TSP.2017.2666775 %0 Conference Paper %B IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) %D 2017 %T Robust clustering of data collected via crowdsourcing %A Alba Pagès-Zamora %A Georgios Giannakis %A R. López-Valcarce %A Pere Gimenez-Febrer %K winter %K wsn %X

Crowdsourcing approaches rely on the collection of multiple individuals to solve problems that require analysis of large data sets in a timely accurate manner. The inexperience of participants or annotators motivates well robust techniques. Focusing on clustering setups, the data provided by all annotators is suitably modeled here as a mixture of Gaussian components plus a uniformly distributed random variable to capture outliers. The proposed algorithm is based on the expectation-maximization algorithm and allows for soft assignments of data to clusters, to rate annotators according to their performance, and to estimate the number of Gaussian components in the non-Gaussian/Gaussian mixture model, in a jointly manner.

%B IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) %C New Orleans %P 4014 - 4018 %8 03/2017 %G eng %R 10.1109/ICASSP.2017.7952910 %0 Conference Paper %B IEEE Int. Conference on Acoustics, Speech and Signal Processing (ICASSP) %D 2015 %T Spectrum Cartography using quantized observations %A Daniel Romero %A Seung-Jun Kim %A R. López-Valcarce %A Georgios Giannakis %K cognitive radio %K compass %K compressed sensing %K spectrum sensing %X
This work proposes a spectrum cartography algorithm used for learning the power spectrum distribution over a wide frequency band across a given geographic area. Motivated by low-complexity sensing hardware and stringent communication constraints, compressed and quantized measurements are considered. Setting out from a nonparametric regression framework, it is shown that a sensible approach leads to a support vector machine formulation. The simulated tests verify that accurate spectrum maps can be constructed using a simple sensing architecture with significant savings in the feedback.
%B IEEE Int. Conference on Acoustics, Speech and Signal Processing (ICASSP) %I IEEE %C Brisbane, Australia %8 04/2015 %G eng