DIGITAL MODULATION CLASSIFICATION

Abstract

The modulation recognition plays an important role in various civilian and military applications. In blind environments, an effective recognition algorithm is needed which should be able to discriminate between digital communication signals. Many researches have worked on the problem of modulation recognition of digitally modulated signals. Many of the algorithms developed need a priori information of a few parameters and thus crops up a space for a class of algorithms that are independent of any information of the signal. This novelty presents a strong case for a leap forward towards blind recognition of the modulation. A feature based algorithm extracts discriminatory features for data representation amongst different modulation schemes and then makes a decision based on thresholds. The first of the algorithm discussed is essentially a sequence of steps which employs statistical parameters such as fourth order cumulants, variance of the centered normalized signal amplitude, instantaneous phase, the zero-crossing sequence shape (ZCSS) and power spectral density as features for classification by employing opportunely set thresholds. The other algorithm approaches the classification problem as multiresolution analysis and comes up with the idea of employing wavelet transform of the signal. The variance of the magnitude of the signal wavelet transform (WT) after peak removal is compared against a set threshold and classification decision is made. It is prudent to mention that a digitally modulated waveform is cyclostationary and exhibits transients whenever a symbol undergoes a transition. The difference is thus exploited for modulation classification.