Multistage Clustering Based Automatic Modulation Classification

Abstract

Automatic modulation classification (AMC) is the problem of identifying the modulation type of a given radio frequency (RF) signal. This operation is one of the key steps in a cognitive radio based spectrum sharing communication network. It is known that the optimal classification algorithms for AMC are computationally intensive which renders real-time implementation almost impossible. In this paper, we propose a practical AMC algorithm that employs multiple stages of clustering to identify the modulation type of the received RF signal. Here, we consider the communication signals to be modulated using the most common digital modulation types: phase shift keying (PSK) or quadrature amplitude modulation (QAM). First, we present a novel algorithm that performs multiple stages of clustering to identify the clusters present in the received data and classifies it to one of the several possible modulation types. Second, we validate our proposed algorithm through practical implementation using software defined radios (SDR). Our results show that the proposed multistage clustering based AMC algorithm works well in practical conditions.