Cyclic Modulation Recognition

A modulation recognition method that uses theoretical cyclic cumulant values as discriminating features. Higher-order cyclic cumulants are inherently insensitive to additive Gaussian noise and carrier phase rotation, making them robust identifiers. A hierarchical decision tree compares extracted cumulant magnitudes against known theoretical values for BPSK, QPSK, 16-QAM, and 64-QAM to determine the modulation type with high confidence.

A one-dimensional projection of the Spectral Correlation Function (SCF) magnitude along the cyclic frequency axis. The CDP serves as a compact feature vector that captures the unique periodicity signature of each modulation scheme. Recognition is performed by cross-correlating the extracted CDP against a library of theoretical profiles, with the highest correlation score indicating the modulation type.

Each digital modulation scheme exhibits unique cyclic frequencies tied to its physical parameters:

• BPSK: Strong cyclostationarity at 2fc ± Rsym due to the squaring operation
• QPSK/OQPSK: Features appear at 4fc after quadrupling to remove phase modulation
• MSK/GMSK: Cycle frequencies at 2fc ± Rsym/2 from the frequency-shift structure
• OFDM: Cyclic prefix induces a correlation peak at the symbol rate

A structured classification approach that sequentially tests for the presence of specific cyclic features to narrow down modulation candidates. The tree first distinguishes between single-carrier and multi-carrier schemes using cyclic prefix detection, then branches to identify the exact constellation order by evaluating higher-order cyclic cumulant magnitudes at candidate cycle frequencies.

Cyclic modulation recognition exploits the fact that stationary Gaussian noise has no cyclostationary content at non-zero cycle frequencies. By evaluating the spectral coherence or cyclic cumulants at α ≠ 0, the classifier effectively rejects noise contributions. This property enables reliable modulation identification even at negative signal-to-noise ratios (SNR) where conventional constellation-based methods fail completely.

Beyond modulation type, cyclic analysis enables blind estimation of key signal parameters without prior knowledge:

• Symbol rate: Extracted as the fundamental cyclic frequency from the SCF
• Carrier frequency offset: Identified from the shift in cycle frequency peaks
• Pulse shaping factor: Estimated from the roll-off characteristics in the cyclic spectrum These parameters feed directly into demodulation chains for signal exploitation.

A robust modulation recognition method that uses higher-order cyclic cumulants as discriminating features. Unlike second-order methods, cyclic cumulants are insensitive to additive Gaussian noise and phase rotation, making them ideal for low-SNR environments.

• Classifies QAM, PSK, and APSK constellations
• Exploits theoretical cumulant values at specific cyclic frequencies
• Resilient to carrier frequency offset and phase jitter

A one-dimensional projection of the spectral correlation function magnitude along the cyclic frequency axis. The CDP compresses the two-dimensional cyclostationary signature into a compact feature vector suitable for machine learning classifiers.

• Reduces dimensionality for real-time processing
• Captures symbol rate, carrier offset, and frame rate peaks
• Used as input to support vector machines and neural networks

A representation displaying the cyclic frequency content of a signal's envelope or instantaneous frequency. This technique reveals modulation-specific periodicities by analyzing the amplitude and phase trajectories of complex constellations.

• Distinguishes FSK from PSK by envelope periodicity
• Identifies OQPSK via staggered phase transitions
• Robust to frequency-selective fading

The FFT Accumulation Method is a computationally efficient channelized algorithm for estimating the spectral correlation function. It decimates the signal into narrowband frequency bins using a sliding FFT, enabling real-time cyclostationary feature extraction.

• Balances temporal and spectral resolution
• Suitable for FPGA and GPU implementation
• Enables blind symbol rate estimation

The process of building a structured, machine-readable representation of a signal's cyclostationary signature. Feature vectors are typically formed by sampling the spectral coherence or cyclic domain profile at key cyclic frequencies corresponding to known modulation parameters.

• Selects peaks at alpha = symbol rate, 2x carrier offset
• Normalizes for scale-invariant classification
• Feeds into convolutional neural networks for deep learning recognition

A statistical hypothesis test that determines whether a signal exhibits cyclostationarity at a candidate cyclic frequency. It evaluates the consistency of the cyclic autocorrelation estimate against a null hypothesis of stationarity.

• Validates detected cyclic frequencies before classification
• Reduces false positives in blind recognition systems
• Uses chi-squared or F-distribution test statistics