Also known as: Norbert Wiener
Norbert Wiener (1894–1964) was an American mathematician who founded the field of cybernetics and derived the Wiener filter, the linear filter that minimises the mean-square error between a desired signal and an estimate formed from a noisy observation.12 His work turned the intuitive idea of “cleaning up” a signal into a precise optimisation problem, and the optimal-estimation viewpoint he introduced underpins much of modern communications and control theory.
Life and work
Wiener was a child prodigy who earned his PhD from Harvard at eighteen and spent most of his career at MIT. He made deep contributions to pure mathematics — Brownian motion, harmonic analysis, and the Wiener–Khinchin theorem relating a signal’s autocorrelation to its power spectrum — before turning, during and after World War II, to the practical problem of predicting and smoothing noisy time series.1 A wartime project on automatic fire-control for anti-aircraft guns, where a gun must be aimed at where an evading aircraft will be, crystallised his thinking about prediction, feedback, and noise.
Contribution
Two strands of Wiener’s work are central to signal processing.
The first is the Wiener filter. Given a desired signal corrupted by additive noise, Wiener asked which linear, time-invariant filter produces the estimate with the smallest average squared error. The answer, expressed through the Wiener–Hopf equation, is a filter whose frequency response weights each band by the ratio of signal power to total (signal-plus-noise) power at that frequency: bands where the signal dominates pass almost untouched, bands buried in noise are suppressed.2 This minimum-mean-square-error (MMSE) criterion is the optimality standard against which later estimators are measured, and the matched filter can be viewed as its special case when the goal is detection rather than waveform recovery.
The second is cybernetics, the study of control and communication in animals and machines, which he named in his 1948 book of that title. Cybernetics framed feedback, information, and regulation as one subject spanning engineering and biology, and helped put the concept of information on an equal footing with energy and matter in the sciences.
Legacy
The Wiener filter is the historical root of statistical, or “optimal”, filtering. Its central limitation — it assumes the signal and noise are stationary, so a single fixed filter serves for all time — was lifted a decade later by Rudolf Kalman, whose Kalman filter recasts the same MMSE goal in a recursive, state-space form that tracks changing signals in real time. When the signal statistics are unknown or drift, an adaptive filter approximates the Wiener solution by learning its coefficients from the data; the LMS algorithm is a stochastic-gradient method that descends toward the Wiener optimum sample by sample. Every one of these traces its optimality criterion back to Wiener.
Relevance to SDR
MMSE thinking is everywhere in a software radio. Channel equalisers that undo multipath, noise-reduction stages, and interference cancellers are all, at heart, attempts to build a Wiener filter for a channel whose statistics must be estimated on the fly — which is exactly what adaptive and Kalman-based methods do. The Wiener–Khinchin relationship also justifies the everyday practice of estimating a signal’s power spectrum by Fourier-transforming its autocorrelation. GopherTrunk does not implement a general Wiener filter by name, but its symbol-timing and carrier-recovery loops and its equalisation stages all rest on the estimation-in-noise foundation Wiener laid.
Sources
-
Norbert Wiener — Wikipedia, for biography, cybernetics, and the Wiener–Khinchin theorem. ↩ ↩2
-
Wiener filter — Wikipedia, for the MMSE criterion and the Wiener–Hopf solution. ↩ ↩2