Publicado em: 21/08/2014

Dissertação de Mestrado em Redes de Computadores

UNIVERSIDADE FEDERAL DO RIO GRANDE DO SUL

INSTITUTO DE INFORMÁTICA

PROGRAMA DE PÓS-GRADUAÇÃO EM COMPUTAÇÃO

———————————————————

DEFESA DE DISSERTAÇÃO DE MESTRADO

Aluno: Maicon Kist

Orientador: Prof. Dr. Juergen Rochol

Título: Adaptive Threshold Architecture for Spectrum Sensing

Linha de Pesquisa: Redes de Computadores e Comunicação de Dados

Data: 28/08/14

Hora: 13:30h

Local: Prédio 43412 – Auditório Inferior, Instituto de Informática

Banca Examinadora:

Prof. Dr. Alberto Egon Schaeffer Filho (UFRGS)

Prof. Dr. José Ferreira de Rezende (UFRJ)

Profa. Dra. Liane Margarida Rockenbach Tarouco (UFRGS)

Prof. Dr. Luciano Paschoal Gaspary (UFRGS)

Presidente da Banca: Prof. Dr. Juergen Rochol

Resumo:

The current spectrum allocation policy comprises licensing the usage of channels of the radio spectrum and ensuring that licensed users have exclusive access to these channels. Through this policy, the best channels for short and long wireless communications were already allocated. Thus, it has become exceedingly hard to find vacant radio channels to either deploy new wireless services or to enhance existing ones. However, recent measurements of the radio spectrum showed that some allocated channels are rarely utilized in certain geographical areas. The relatively low utilization of some radio channels made the governmental agencies consider a new spectrum access policy. In this new access policy, an unlicensed user can temporarily access underutilized radio channels, with the constraint of not interfering with the transmission of any licensed user. Thus, the unlicensed user must analyze the radio channel before accessing it, with the objective of guaranteeing that no licensed user is transmitting. The analysis of radio channels is made through spectrum sensing solutions. The major drawback of current spectrum sensing solutions is the use of static parameters to detect the transmission of a licensed user. The usage of static parameters is a drawback because the spectrum sensing may encounter different noise and interference levels during the channel analysis. In this context, machine learning algorithms can be employed to dynamically adapt the detection parameters used in the spectrum sensing. In this dissertation we propose the Adaptive Threshold architecture (ATA) for spectrum sensing. This architecture employs machine learning algorithms to adapt the detection parameters of the spectrum sensing in real time. Furthermore, a prototype of ATA was developed and evaluated in an experimental radio environment based in the IEEE 802.22 standard. The results  show that ATA achieves a better performance than current spectrum sensing solutions in terms of the accuracy in detecting the licensed user and in the time required to analyze the radio channel.

Palavras-chave:  spectrum sensing, machine learning, dynamic spectrum access, wireless communication

_____________

Divulgação PPGC