A Power-Weighted Noncentral Complex Gaussian Distribution

Researcher Toru Nakashika has introduced a novel probabilistic model for complex-valued data called the Power-Weighted Noncentral Complex Gaussian Distribution. Published on arXiv, this work addresses limitations in traditional signal processing models where standard complex Gaussian distributions struggle to represent diverse amplitude characteristics of individual source signals. The key innovation is a formulation that operates directly on the complex plane while retaining a higher-dimensional interpretation, preserving the geometric structure inherent in complex-valued observations that many hyperspherical models overlook.

The model's core mechanism is a single shape parameter that enables continuous control over distributional geometry. This parameter creates nonlinear phase diffusion, allowing the distribution to shift from arc-shaped diffusion along the phase direction to concentration of probability mass toward the origin. This flexibility means researchers and engineers can model a wider range of real-world signal behaviors with one unified framework rather than switching between different specialized distributions.

Experimental validation on speech power spectra demonstrated the model's practical superiority, consistently outperforming conventional distributions in terms of log-likelihood. The derived amplitude and power distributions provide a unified framework that encompasses several widely used distributions in signal modeling, including the Rice, Nakagami, and gamma distributions. This represents a significant theoretical advancement that bridges gaps between different modeling approaches while offering improved empirical performance for applications like audio processing and communications.