This article presents an in-depth study of AMR on LB and ML approaches. LB approaches provided better accuracy, but poor handling of more unknown parameters made them to be rarely used. ML approaches used different classifiers (KNN, SVM, DT) shown good results on classification accuracy. ML used different ways to generate the features like cumulants, moments and statistical features. It achieved promising results in different operating conditions with different SNRs making them highly recommended in realistic environment. ML approaches cannot handle big data sets resulting in performance degradation.
展开▼
