AUTOMATED MODULATION CLASSIFICATION SYSTEM FOR SOFTWARE DEFINED RADIO

Abstract

An Automatic Modulation Classification (AMC) system for Software Defined Radio (SDR) is presented in this study. Initially, the generated signals are modulated using different modulation techniques. Then, noise is added to the generated signals by using Additive White Gaussian Noise (AWGN). The noise added signal is used for further process to extract features and classification. The system uses Discrete Wavelet Transform (DWT) to analyze the signal that produces lower and higher frequency sub-bands. The Independent Component Analysis (ICA) is employed on lower frequency sub-band for dimensionality reduction. Finally, the classification is made by Pulse Coupled Neural Network (PCNN). The system uses three different digital modulation schemes; Phase Shift Keying (PSK), Quadrature Amplitude Modulation (QAM), and Differential PSK (DPSK). The results show the DWT, ICA and PCNN based AMC system provides promising results under various noise densities.

Keywords

• automated modulation classification, machine learning algorithms, dimensionality reduction techniques, software defined radio