Degree of Cyclostationarity

The DCS metric is fundamentally a normalized ratio that quantifies the energy contained in a signal's cyclostationary components relative to its total energy. It is computed by integrating the Spectral Correlation Function (SCF) over all cycle frequencies and normalizing by the signal's total power. The result is a single value between 0 and 1, where 0 indicates a purely stationary process and 1 indicates a perfectly cyclostationary signal. This normalization makes the metric independent of absolute signal power, allowing for robust comparison across different signal-to-noise ratio (SNR) conditions.

Different modulation schemes exhibit distinct DCS values, making the metric a powerful feature for automatic modulation classification (AMC). For example:

• BPSK signals typically show a high DCS due to strong cyclostationarity at twice the carrier frequency and symbol rate.
• QPSK signals have a lower DCS than BPSK because their balanced constellation suppresses some second-order cyclic features.
• Higher-order QAM (e.g., 64QAM) signals approach a Gaussian distribution, yielding a DCS closer to 0.
• OFDM signals exhibit a unique DCS profile due to the cyclostationarity induced by the cyclic prefix.

A key advantage of the DCS metric is its inherent immunity to stationary Gaussian noise. Because stationary noise has no cyclic features, its contribution to the cyclic spectrum is zero at non-zero cycle frequencies. The DCS metric explicitly measures the ratio of cyclostationary energy to total energy, meaning that additive white Gaussian noise (AWGN) only increases the denominator, driving the DCS toward zero in a predictable manner. This property makes DCS-based detectors and classifiers significantly more robust in low-SNR environments compared to energy-based methods.

In practice, the DCS is estimated from a finite data record using a computed Spectral Correlation Function (SCF). The estimation process involves:

• Computing the SCF using efficient algorithms like the FFT Accumulation Method (FAM) or the Strip Spectral Correlation Analyzer (SSCA).
• Integrating the magnitude-squared SCF over the entire bifrequency plane to capture total cyclostationary energy.
• Normalizing by the integrated power spectral density. The choice of estimation algorithm introduces a bias-variance trade-off that must be managed for reliable DCS computation.

The DCS metric serves as a test statistic for spectrum sensing in cognitive radio. A binary hypothesis test can be formulated:

• Null Hypothesis (H0): The band contains only noise, yielding a DCS near 0.
• Alternative Hypothesis (H1): A modulated signal is present, yielding a DCS above a predetermined threshold. This approach outperforms energy detection in low-SNR regimes and can distinguish between different primary user signals, enabling more intelligent dynamic spectrum access strategies.

The DCS is closely related to the Spectral Coherence Function, which provides a frequency-resolved measure of cyclostationarity. While the spectral coherence function gives a value for every pair of spectral frequency and cycle frequency, the DCS collapses this multidimensional information into a single global metric. This aggregation trades spatial resolution for simplicity, making DCS ideal for applications requiring a quick, high-level assessment of signal cyclostationarity, such as coarse modulation screening or real-time spectrum monitoring dashboards.