Cyclic Periodogram

The cyclic periodogram is defined as the product of two frequency-shifted, finite-time Fourier transforms of a signal. For a finite-length signal segment x(n), it is computed as S_x^α(f) = (1/N) * X_T(f + α/2) * conj(X_T(f - α/2)), where X_T(f) is the complex Fourier transform of the segment, α is the cyclic frequency, and f is the spectral frequency. This direct multiplication reveals the correlation between spectral components separated by the cyclic frequency α.

A critical property of the cyclic periodogram is its statistical inconsistency. As the data observation length N increases, the variance of the estimate does not converge to zero. This means the estimate remains noisy regardless of how much data is processed. This inconsistency arises because the spectral resolution of the periodogram is fixed by the segment length, preventing the estimator from averaging out random fluctuations effectively.

The cyclic periodogram is an asymptotically unbiased but inconsistent estimator of the ideal Spectral Correlation Function (SCF). The SCF is the true, theoretical limit of the spectral correlation density. The periodogram's expected value converges to the SCF, but its variance does not diminish. This relationship necessitates the use of frequency or time smoothing techniques to transform the raw periodogram into a consistent SCF estimate.

The direct computation of the cyclic periodogram for all points on the bifrequency plane (f, α) is computationally intensive, scaling with O(N²) for an N-point FFT. This high cost makes it impractical for real-time applications. Efficient algorithms like the FFT Accumulation Method (FAM) and the Strip Spectral Correlation Analyzer (SSCA) were developed specifically to compute the cyclic periodogram (or its smoothed variants) with significantly reduced complexity by trading off resolution for computational speed.

The finite-time Fourier transform used in the cyclic periodogram introduces spectral leakage. This manifests as energy from a strong spectral component spreading into adjacent frequency bins. In cyclostationary analysis, a related phenomenon called cycle leakage occurs, where the cyclic periodogram at a specific cyclic frequency α is contaminated by energy from strong features at nearby cyclic frequencies. Proper windowing and smoothing are essential to mitigate these artifacts.

Despite its inconsistency, the cyclic periodogram serves as the raw input for feature extraction in Automatic Modulation Classification (AMC). The distinct patterns of peaks in the periodogram's bifrequency plane form a unique cyclic signature for each modulation scheme. For example, a BPSK signal exhibits strong cyclic features at α = 2f_c (twice the carrier frequency) and α = 2f_c ± R_s (where R_s is the symbol rate), which are directly visible in the periodogram before any smoothing is applied.

The Spectral Correlation Function is the theoretical two-dimensional transform that the cyclic periodogram attempts to estimate. It measures the correlation between frequency-shifted versions of a signal at spectral frequency f and cyclic frequency α. Unlike the periodogram, the SCF is a deterministic function that reveals hidden periodicities in a signal's spectral content, making it a powerful tool for blind modulation classification.

A normalized version of the Spectral Correlation Function that quantifies the degree of correlation between frequency-shifted signal components on a scale from zero to one. This normalization removes the influence of signal power, making it a robust feature for modulation recognition in varying channel conditions. A value near 1 indicates strong cyclostationarity at a given cyclic frequency.

The FFT Accumulation Method (FAM) is a computationally efficient algorithm for estimating the spectral correlation function. It uses a channelizer and short-time FFTs to compute the cyclic periodogram and then averages these estimates over time. FAM is the practical, consistent estimator that overcomes the cyclic periodogram's inherent variance, making it suitable for real-time spectrum monitoring.

The Strip Spectral Correlation Analyzer (SSCA) is an alternative to the FAM algorithm for estimating spectral correlation. It offers different trade-offs between computational complexity and spectral resolution. SSCA processes data in strips, making it highly parallelizable and well-suited for FPGA-based implementations in real-time signal intelligence systems.

The cyclic frequency, denoted by α, is the fundamental parameter that quantifies the periodicity of a signal's statistical moments. For a digitally modulated signal, key cyclic frequencies include the symbol rate and multiples of the carrier frequency offset. The cyclic periodogram computes correlation at specific α values, revealing the unique cyclic signature of each modulation scheme.

The time-domain counterpart to the cyclic spectrum. This function measures the correlation of a signal with a frequency-shifted and conjugated version of itself, parameterized by a cyclic frequency α. The Fourier transform of this function yields the cyclic spectrum. It is the fundamental building block for understanding how second-order cyclostationarity manifests in modulated signals.