Database Assisted Automatic Modulation Classification Using Sequential Minimal Optimization

TL;DR

This paper introduces a database-assisted, SMO-based algorithm for automatic modulation classification in cognitive radio systems, achieving high accuracy and efficiency in identifying unknown signals.

Contribution

It presents a novel database approach combined with SMO for faster, accurate modulation classification, especially in complex scenarios with multiple modulation schemes.

Findings

Over 99% accuracy at 15 dB SNR

Over 95% accuracy at 5 dB SNR

Significant reduction in computational complexity

Abstract

In this paper, we have proposed a novel algorithm for identifying the modulation scheme of an unknown incoming signal in order to mitigate the interference with primary user in Cognitive Radio systems, which is facilitated by using Automatic Modulation Classification (AMC) at the front end of Software Defined Radio (SDR). In this study, we used computer simulations of analog and digital modulations belonging to eleven classes. Spectral based features have been used as input features for Sequential Minimal Optimization (SMO). These features of primary users are stored in the database, then it matches the unknown signal's features with those in the database. Built upon recently proposed AMC, our new database approach inherits the benefits of SMO based approach and makes it much more time efficient in classifying an unknown signal, especially in the case of multiple modulation schemes to overcome the issue of intense computations in constructing features. In various applications, primary users own frequent wireless transmissions having limited their feature size and save few computations. The SMO based classification methodology proves to be over 99 \% accurate for SNR of 15 dB and accuracy of classification is over 95 \% for low SNRs around 5dB.