PROTOTIPE MODULATION CLASSIFIERBERBASIS MACHINE LEARNINGMENGGUNAKAN SOFTWARE DEFINED RADIO

• Main Author: Eldy Rashad, Farras
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/55329
• Institution: Institut Teknologi Bandung

In a modern communication system, a wireless receiver with Automatic Modulation Classification (AMC) ability has a very vast application potential. A combination of Machine Learning (ML) and AMC will increase the system performance significantly. The flexible Software Defined Radio (SDR) plays an important role in combining ML with AMC. ML traits which improve classification performance through training processes will benefit greatly by the usage of SDR. In this project, the three concepts are implemented for ADALM-PLUTO SDR. For the developed prototype, modulation types are limited to BPSK, QPSK, 8-PSK, 16-QAM, and 8- PAM. A successful classification requires a selection of features, therefore in this project 6 spectral, high order statistics, and wavelet transform based features are selected. This project also uses a QAM synchronization scheme to mitigate the effects phase and frequency shifts to the feature extraction process. With the help of Support Vector Machine (SVM), we created a classification model for SDR based QAM Receiver with 91.4% accuracy. The Classification model is implemented for the ADALM-PLUTO SDR.