Trispectrum

The trispectrum uniquely identifies cubic phase coupling—non-linear interactions where three frequency components combine to generate a fourth at their sum or difference. This phenomenon arises from third-order non-linearities in transmitter power amplifiers and mixers. Unlike the bispectrum, which detects only quadratic coupling, the trispectrum reveals more subtle, higher-order hardware impairments that serve as distinctive, unclonable device signatures.

A defining property of the trispectrum is its theoretical immunity to Gaussian noise. All cumulants of order greater than two are identically zero for Gaussian processes. By operating in the fourth-order cumulant domain, the trispectrum completely suppresses additive white Gaussian noise, enabling the extraction of non-Gaussian signal features that would otherwise be buried below the noise floor. This makes it invaluable for low-SNR emitter identification scenarios.

The trispectrum is a function of three independent frequency variables: T(f₁, f₂, f₃). This three-dimensional manifold captures the full fourth-order spectral correlation structure of a signal. Key regions of interest include:

• Stationary manifolds where f₁ + f₂ + f₃ = 0
• Diagonal slices for dimensionality reduction
• Principal support regions where cubic coupling energy concentrates Each region reveals different aspects of transmitter non-linearity.

Unlike the power spectrum, which discards all phase information, the trispectrum preserves both magnitude and phase relationships between frequency components. The trispectral phase contains critical information about the shape and symmetry of a transmitter's non-linear transfer function. This phase preservation enables more discriminative fingerprinting than magnitude-only spectral methods, as phase distortions are highly specific to individual hardware chains.

The trispectrum is formally defined as the three-dimensional Fourier transform of the fourth-order cumulant sequence. This relationship provides the mathematical foundation for trispectral estimation:

• Indirect method: Estimate cumulants from time-domain data, then apply multi-dimensional FFT
• Direct method: Average segmented periodograms in the frequency domain
• Parametric method: Fit autoregressive models and compute the trispectrum analytically The cumulant-trispectrum duality ensures consistent statistical interpretation.

The trispectrum's three-dimensional nature introduces significant computational burden—O(N³) complexity for N frequency bins. Practical implementations employ:

• Diagonal slice extraction to reduce to 1D or 2D representations
• Integrated trispectrum along radial paths for compact feature vectors
• Principal component analysis on the trispectral tensor
• Region-of-interest masking based on known coupling physics These optimizations make real-time trispectral analysis feasible on modern SDR platforms.