Method to reduce the signal‐to‐noise ratio required for modulation recognition based on logarithmic properties

Abstract

Here, the authors present a novel, simple, and effective way of improving additive white Gaussian noise resistance and reducing the signal-to-noise ratio (SNR) required for modulation recognition. Working on the theoretical basis that the ratio of two logarithmic functions with the same variable is a constant, the authors selected QPSK, 16QAM, and 64QAM for investigation. For each of these, the authors constructed distribution curves of higher-order cumulants, using SNR as the variable, and examined how they might work as logarithmic curves. First, their logarithmic similarities were measured. Then their former features were divided by logarithmic functions to construct new features whose distribution curves were more parallel to the threshold line. Finally, an algorithm for classifying the selected modulation formats was designed whose computational complexity was then compared with that of adaptive-threshold classification algorithm, and the recognition rate was assessed statistically. For the purposes of validation, the algorithm was tested experimentally using actual signals. The experiment confirmed that the new logarithmic features with fixed thresholds were able to maintain an efficient recognition rate of 80% when SNR was reduced from 11 to 6 dB, and suffered less computational complexity than traditional cumulants with adaptive thresholds which were achieved by support vector machine.