Adaptive threshold spectrum sensing based on Expectation Maximization algorithm

In this article we address a novel method for spectrum sensing, based on the Expectation Maximization algorithm applied to the histogram of the moving average signal power. The method enables the estimation of the number of active users in a given frequency band, the power received from each user, the occupied time slots and the front-end noise floor. The proposed approach takes advantage of the statistical properties of the averaging estimator output, which allows to model the received estimated power as a Gaussian mixture. This model represents the distributions of the users transmitted signal power as well as the system noise floor. Moreover, the Gaussian with the lowest mean that is related with the noise floor, can be used to estimate an adaptive threshold for a constant false alarm rate detector. Finally, the method was validated in a Wi-Fi experimental setup, where real-world data was acquired with a software defined radio.