In this work, a novel multiband spectrum sensing technique is implemented in the context of cognitive radios. This technique is based on multiresolution analysis (wavelets), machine learning, and the Higuchi fractal dimension. The theoretical contribution was developed before by the authors; however, it has never been tested in a real-time scenario. Hence, in this work, it is proposed to link several affordable software-defined radios to sense a wide band of the radioelectric spectrum using this technique. Furthermore, in this real-time implementation, the following are proposed: (i) a module for the elimination of impulsive noise, with which the appearance of sudden changes in the signal is reduced through the detail coefficients of the multiresolution analysis, and (ii) the management of different devices through an application that updates the information of each secondary user every 100 ms. The performance of these linked devices was evaluated with encouraging results: 95% probability of success for signal-to-noise ratio (SNR) values greater than 0 dB and just five samples (mean) in error of the edge detection (start and end) for a primary user transmission.

In this work, three specific machine learning techniques (neural networks, expectation maximization and k-means) are applied to a multiband spectrum sensing technique for cognitive radios. All of them have been used as a classifier using the approximation coefficients from a Multiresolution Analysis in order to detect presence of one or multiple primary users in a wideband spectrum. Methods were tested on simulated and real signals showing a good performance. The results presented of these three methods are effective options for detecting primary user transmission on the multiband spectrum. These methodologies work for 99% of cases under simulated signals of SNR higher than 0 dB and are feasible in the case of real signals.