<![CDATA[This research uses randomly generated digital data to detect modulation types. The types of modulation are tested QPSK, 16-QAM, and 64-QAM.In the characteristic extraction process uses a statistical feature set approach with the mean, varian, kurtosis and skewness, while feature selection uses Multi-Class Support Vector Machine(SVM) with 5 classes in the classification including (i)Not a feature, (ii)Mean, (iii)Varian, (iv)Kurtosis and (v)Skewness.In detecting the type of modulation in this study using Backpropagation Artificial Neural Networks by learning process used the Conjugate Gradient Polak Ribiere algorithm. This study, researcher also compared the learning with Conjugate Gradient Polak Ribiere and learning using Gradient Discent. The results of the comparison learning was 401 training data using Conjugate Gradient Polak Ribiere much better with a higher accurancy value of 86,20% and lower error rate of 13,80% in the iteration to 781, while the Gradient Discent in the same iteration the accuracy rate is 67,83% and the error rate is 32,17%. From the test results there are 4 feature groups that was able to recognize the type of modulation including (i) Mean,Variant,Kurtosis, (ii)Mean, Variant, Skewness, (iii)Variant, Kurtosis, Skewness and (iv)Mean, Kurtosis, Skewness.]]>
Copyrights © 2019
