Matched filter

A matched filter is the optimal way to dig a known signal shape out of noise — the receiver's best possible ear for hearing a faint, expected pulse. The idea is to correlate the incoming noisy waveform against a stored copy of the exact pulse you're looking for. Wherever the incoming wave lines up with the template, the correlation builds to a sharp peak; the random noise, having no such shape, mostly cancels itself out. It's like recognising a friend's specific knock on the door amid a clatter of random sounds.

Mathematically the matched filter is the time-reversed copy of the transmitted pulse, and it is provably the linear filter that maximises SNR at the sampling instant — no other linear receiver can do better against white noise. This is why every serious digital receiver, from a Wi-Fi chip to a deep-space probe to a GPS unit (which correlates against the satellite's known pseudo-random code), is built around a matched filter or its discrete-time cousin, the correlator. It's the quiet workhorse that lets us recover bits buried far below the noise floor.