A Method for Modulation Recognition Based on Entropy Features and Ensemble Algorithm

Ensemble learning is a useful frame algorithm which could improve the performance of weak learners by combining them. It is well known AdaBoost algorithm is one of these successful boosting algorithms. In this paper, we choose it to complete ensemble frame. we compare the performance of three machine learning algorithms including SVM, AdaBoost and decision tree stump based on communication signal modulation scheme to prove the effect of AdaBoost. The AdaBoost algorithm combines decision tree stump and iterates 500 rounds in the training phase. And the result reveals the performance of AdaBoost is proximal to that of SVM. At last, experiment to examine the features' working principle on signals is done. The features can identify 4ASK correctly in all SNRs.