A voting approach of modulation classification for wireless network

Atik Mahabub, Al-Zadid Sultan Bin Habib

Abstract: Radio signal modulation has been a key task for a wireless communication system. In this paper, a novel voting approach has been introduced for the identification of radio signal modulations for wireless network systems using the statistical Machine Learning (ML) based method. Several well-known ML classifiers have been used here before using it for the voting approach. The most effective ML algorithms have been utilized in Ensemble Voting Classifier (EVC). In the end, we found that this voting approach provides over 90% accuracy for the high SNR values for eight different wireless network modulation schemes. This model has also performed better in case of low SNR values and provided negligible loss for each individual modulations compared to the existing methods.