Cognitive network spectrum sensing using the Gaussian mixture model

Cognitive radio (CR) technique shows promise in addressing spectrum shortages caused by rapid technological advancements. However, a key challenge in cognitive networks is the hidden primary node problem, where cognitive users might misjudge the occupied spectrum, causing interference with primary users. To solve this issue, a cooperative sensing method employing machine learning has emerged. Two popular machine-learning approaches in CSS are the Gaussian Mixture Model - GMM and the Support Vector Machine - SVM. GMM, the unsupervised learning method, models spectrum occupancy features using a combination of Gaussian density of distributions, with parameters found in the training phase using training data. In contrast, SVM, a supervised learning technique, constructs a decision surface using a subset of training vectors to distinguish between available and unavailable channels. Its goal is to enhance classification performance by maximizing the margin between training vectors and the decision boundary.