Adaptive Expectation-Maximization Detection of Gaussian Signals in Quantized Systems with Jointly Unknown Statistical Parameters

الملخص

This  paper  addresses  the  challenging  problem  of detecting  Gaussian  signals  with jointly unknown  statistical pa- rameters  (mean,  variance)  in   the  presence   of  additive  white Gaussian  noise  (AWGN)  and  low-resolution  quantization.  We propose  a  novel  detection  scheme  based  on  the  Expectation- Maximization  (EM)  algorithm,  which  iteratively  estimates  the unknown parameters from the quantized observations. The un- quantized received signal is treated as a latent variable, allowing for a tractable derivation of the E-step and M-step. Subsequently, an EM-based Generalized Likelihood Ratio Test (GLRT) detector is  formulated.  The  performance  of  the  proposed  detector  is analyzed  through  simulations,  demonstrating  its  effectiveness in  various   quantization  scenarios  and  comparing  it  against theoretical  benchmarks.  This  work  provides  a  robust  solution for  signal  detection  in  power-constrained  and  high-frequency communication systems, such as those envisioned for 6G, where low-resolution analog-to-digital converters (ADCs) are prevalent