Automatic Modulation Recognition through Clustering Analysis of Constellation Signature

Abstract

Automatic recognition of Digital Modulation schemes is becoming an active research area in many covert operations. Most of the approaches were based on modulated signals component, but the modulation type can be best identified with the use of constellation diagram. The proposed technique in this thesis is able to discriminate between M- QAM, MASK and M-PSK type of modulation scheme at low value of SNR. As constellation points form clusters in I-Q plane, order of the modulation is obtained by estimating the optimized number of clusters in the I-Q plane, which is identified by using Hierarchical clustering algorithm and density based OPTICS clustering algorithm. This thesis also compares the performance of both the clustering algorithm. To identify the domain of modulation between MASK, MPSK and MQAM, least square error has been calculated using linear regression analysis which is least for MASK and it is differentiated from the MPSK and MQAM. Further to identify between the remaining two modulation schemes, number of centroids equal to order calculated earlier has been estimated using k-means clustering algorithm, and with the difference in the absolute value of the maximum and minimum centroid values, MPSK and MQAM are classified. The approach presented is completely unsupervised and performs the task for varying SNR values. The simulation of this method shows high capability for recognition of modulation scheme in the presence of the AWGN noise. The algorithm developed has also been implemented in labVIEW and tested using NI PXIe-5673 (RF transmitter), NI PXI 5661 (RF receiver).