Identifying spectrum usage by unknown systems using experiments in machine learning

Authors:
Nikhil Shetty;Sofie Pollin;Przemysław Pawełczak
Affiliations:
Department of EECS, University of California, Berkeley, CA;Department of EECS, University of California, Berkeley, CA and Interuniversity Micro-electronics Center, Leuven, Belgium;Department of EEMCS, Delft University of Technology, Delft, The Netherlands
Venue:
WCNC'09 Proceedings of the 2009 IEEE conference on Wireless Communications & Networking Conference
Year:
2009

Citing 6
Cited 0

Variational methods for the Dirichlet process

ICML '04 Proceedings of the twenty-first international conference on Machine learning
A framework for clustering evolving data streams

VLDB '03 Proceedings of the 29th international conference on Very large data bases - Volume 29
Wireless device identification with radiometric signatures

Proceedings of the 14th ACM international conference on Mobile computing and networking
Applications of Machine Learning to Cognitive Radio Networks

IEEE Wireless Communications
Cognitive functionality in next generation wireless networks: standardization efforts

IEEE Communications Magazine
Spectrum sensing: A distributed approach for cognitive terminals

IEEE Journal on Selected Areas in Communications

Quantified Score

Hi-index	0.01

Visualization

Abstract

We adopt a machine learning approach towards the problem of identifying wireless systems present in a dynamic radio environment with heterogeneous usage. To classify the wireless systems, we utilize two features that typify spectrum use--center frequency and the frequency spread--and cluster the measurement data in this space. Since the systems are unknown prior to clustering, we use an unsupervised clustering method that uses the Chinese restaurant process implemented using Gibbs sampling. The system identification is divided into two parts: training and online classification. In the training phase, we assign wireless systems present in the surrounding to the clusters while the online classification uses this trained data to perform classification. By means of an extensive measurement campaign, we show that the proposed machine learning process achieves up to 90% correctness in classifying the wireless systems considered here.