Combination Spectrum Sensing Algorithm for Wireless Sensor Network Based on Random Forest

Abstract

In this paper, a multi-conditional spectrum sensing combination algorithm based on random forest is proposed to address the current shortage of spectrum resources in the sensor network. The algorithm combines sensor's velocity, signal energy, the traces, and the average eigenvalue of the covariance matrix as random forest characteristic parameters, which are achieved through the strong multi-classification ability of random forest. To improve the successful rate of spectrum sensing and the utilization rate of the spectrum, we focus on analyzing the selection of parameter in theory as well as the low signal-to-noise ratio caused by channel fading and shadow effect. Meanwhile, the Doppler effective caused by car moving is also our consideration. Under low signal-to-noise ratio, the simulation results show that the proposed algorithm has better detection performance than existing spectrum sensing algorithms.