Blind Digital Modulation Identification for Spatially-Correlated MIMO Systems

A Low Complexity Modulation Classification Algorithm for MIMO Systems

“…Theoretical values of κ (4,0) and κ (4,2) for different unit variance constellations are provided in Table I. For the theoretical values of the corresponding moments, the reader is referred to [1].…”

“…For the theoretical values of the corresponding moments, the reader is referred to [1]. Note that the estimates of κ (4,0) and κ (4,2) are obtained from (2) and (3) with the moments replaced by their corresponding finite sample estimates [6]. Clearly, the cumulant features cannot be directly estimated from r[k] in (1) due to the channel effects, which are required to be compensated for prior to feature estimation.…”

“…is the cross-covariance matrix ofκ (4,0) and κ (4,2) . The mean of the feature vector depends on the modulation type M q , while its covariance matrix additionally depends on the channel matrix H and the noise variance σ 2 w .…”

“…However, both algorithms exhibit a high computational complexity, which severely limits their applicability. The feature-based approach is followed in [4] and [5]. Diverse higher-order signal statistics are used as features in [4] with a neural network-based algorithm that requires prior training and is sensitive to channel estimation errors.…”

“…The feature-based approach is followed in [4] and [5]. Diverse higher-order signal statistics are used as features in [4] with a neural network-based algorithm that requires prior training and is sensitive to channel estimation errors. The algorithm in [5] employs higher-order signal statistics with a sub-optimal criterion of decision.…”