Performance Analysis of Modulation Classification Using Machine learning

Abstract

Automatic modulation classification is used to identify the modulation scheme of the received signal, without prior knowledge of system parameters. In this work, we compare the performance of modulation classification in additive white gaussian noise channel using a conventional method and a deep learning-based method. Firstly, we classified the modulation schemes using a likelihood-based classifier. Another classifier is also implemented by exploiting the estimated probability density function. Next, a feature-based learning technique using a feedforward neural network was executed. We have analyzed this for digital modulation schemes like BPSK, QPSK, and 16-QAM. The performance of each modulation classification technique in different signal-to-noise ratios is tabulated.