A dynamic threshold computing method based on SOM neural network in frequency occupancy analysis

Spectrum occupancy measurement and analysis is a very important aspect in frequency resource planning by the reason of frequency shortage for the increasing wireless systems. It is also useful in spectrum sensing in the context of cognitive radio, especially in unlicensed band. Given that, a dynamic decision threshold computing method based on SOM Neural Network in spectrum occupancy analysis is proposed in this paper. This method is designed to cluster sample data into different spaces by network self-learning without knowing the statistics of the measurement data. Simulation with pre-set data is given, and the results show the accuracy and the robustness of the proposed method. This dynamic method is validated with the data derived from the spectrum measurements of 806 - 960 MHz band in Beijing.