Logo image
Back
Spectrum awareness using Bayesian nonparametric pattern recognition
Dissertation   Open access

Spectrum awareness using Bayesian nonparametric pattern recognition

Gabriel Ford
Doctor of Philosophy (Ph.D.), Drexel University
May 2018
DOI:
https://doi.org/10.17918/D8PH3X
pdf
Ford_Gabriel_20183.16 MBDownloadView
PDF Open Access Open Access (License Unspecified)

Abstract

Electrical engineering Nonparametric signal detection Bayesian statistical decision theory Pattern recognition systems Machine Learning
To accommodate an increasing demand for scarce spectrum resources, dynamic spectrum access (DSA) opens portions of the spectrum currently dedicated to licensed primary users, for access by unlicensed cognitive radio secondary users. A key requirement is that the secondary users do not interfere significantly with the primary users. Previous DSA research focuses on sense-and-avoid spectrum sharing strategies. The secondary user occupies a primary user channel only when it is not in use, and vacates immediately upon detecting primary user transmissions. In this setting, spectrum sensing is intended to provide little information aside from instantaneous primary user spectral-temporal occupancy. However, we anticipate that more adaptive and intelligent spectrum sharing strategies will require more advanced sensing capabilities allowing a secondary user to infer the primary user higher-layer protocol state and behavior. This type of feedback about the primary user will enable a secondary user to optimize its own spectrum access while also monitoring for potential impact to the primary user. In this study, we explore Bayesian nonparametric pattern recognition as a tool for informing intelligent secondary-user DSA strategies. We present a framework for learning and inferring primary user protocol state at the application and MAC layers from simple low-level energy detector features. We demonstrate, using a physical wireless network testbed, how this approach discovers actual primary user application layer protocol states and also detects anomalous primary user behavior caused by secondary user interference. We then extend the spectrum awareness framework to handle several simultaneous primary/secondary traffic flows, of potentially different types, multiplexed together onto a single wireless broadcast medium. We use the multi-flow framework to infer the application-layer states of the interleaved flows directly from observations of the aggregate traffic. In this process we circumvent deinterleaving the transmissions of the component flows, a particularly difficult task in cognitive radio environments with parameter-agile transmitters. We demonstrate the performance of the resulting technique on a network scenario with multiple simultaneous flows carrying different application layer traffic types, both in emulation and on an over-the-air hardware testbed.

Metrics

79 File views/ downloads
18 Record Views

Details

Logo image