Statistical Digital Signal Processing and Modeling — Hayes

Monson H. Hayes (Wiley)

The random-signals and modeling half of DSP. Discrete-time random processes, autocorrelation and power spectral density, signal modeling (Padé, Prony, autoregressive/moving-average), the Levinson–Durbin recursion, Wiener filtering, adaptive filters (LMS/RLS), and spectrum estimation (periodogram, Welch, parametric methods). This is the statistical foundation under noise-floor measurement, PSD estimation, adaptive noise cancellation, and the classical-signal-modeling view that on-device ML extends.

Deeper reading for Course 1 — Lessons 39–40, the statistical-signal-processing lessons. They build on the probability of Parts III–IV and turn it into random-process theory, signal modeling, Wiener filtering, spectrum estimation, and adaptive filters — including the periodogram bias and variance that put error bars on a measured spectrum. Every chapter carries a problem set plus MATLAB computer exercises; problems are worked by hand.

Used in Course 2 — random-signal theory behind the noise-floor / PSD lab, adaptive filtering, and the statistical-modeling framing for the edge-ML signal-processing module.

Chapters

Notes and worked problems added as I work through each chapter.