Loading...

Effective Implementation of Wide-band Spectrum Sensing

Golvaei, Mehran | 2018

678 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 52245 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Shabany, Mahdi; Fakharzadeh, Mohammad
  7. Abstract:
  8. Ever increasing demand for higher data rate in wireless communication in the face of limited or underutilized spectral resources has motivated the introduction of cognitive radio for dynamic access to spectrum. In dynamic spectrum access a new type of users called secondary users measure the spectrum to see if it is occupied by licensed users (primary users or PU). When channel is empty secondary users can use it to transmit signal. This approach is called spectrum sensing. Hidden PU problem can severely defect detection ability of non-cooperativ spectrum sensing systems. Cooperative spectrum sensing (CSS) uses spatial diversity of spectrum sensors to tackle this problem. There are two kinds of decision making approaches in CSS systems. Hard decision making by which channel state is deiced based on sensors individual decisions and Soft decision by which channel state is decided based on spectrum sensors measurements. In soft decision sensor measurments together make feature vectors and this vectors must be classified to channel-empty or channel-occupied classes. This motivates us to use machine learning algorithms to make such decision. In literature it has been shown that SVM-Linear has the best performance among all machine learning algorithms. Cooperative spectrum sensing systems originally are narrow-band systems or implement wide-band sensing using sequential sensing. In this dissertation a novel cooperative wide band spectrum sensing is introduced based on multi band joint detection idea in which we present a fast decision making algorithm which keeps SVM-Linear detection ability but increases decision making speed by decreasing computational complexity so that, in test conditions, training time falls from 100 milliseconds to 5 milliseconds and test time falls from 8 milliseconds to 1 millisecond
  9. Keywords:
  10. Cognitive Radio ; Wide-band Spectrum Sensing ; Cooperative Spectrum Sensing ; Machine Learning

 Digital Object List

 Bookmark

No TOC