A detection principle that for a known deterministic signal in additive white Gaussian noise prescribes a linear filter whose impulse response is the time‑reversed complex conjugate of the known signal: this filter maximizes the output signal‑to‑noise ratio among all linear filters and implements the maximum‑likelihood detector for that model.

Principle

For additive white Gaussian noise and a known waveform, correlating the received waveform with the matched template (time‑reversed conjugate) maximizes output SNR and yields the optimal linear detector; receiver implementations therefore use matched filtering for optimal detection under these assumptions.

Demonstration

Illustrative scenario → A receiver must detect presence of a known pulse s(t) in AWGN. Recognition → Construct filter h(t)=s*(T−t) where T aligns detection time. Action → Filter the received signal and sample the filter output at T. Consequence → The sampled output provides the maximum achievable SNR for linear filtering, yielding the best detection performance under the AWGN model.

Misapplication

Using the matched filter unchanged in colored noise or in presence of structured interference without whitening: applying the template directly will not maximize SNR and can give suboptimal or misleading detection statistics.

Consequence

Gives a concrete receiver architecture (correlator or matched filter) and performance benchmark; deviations from the AWGN assumptions require preprocessing (whitening), generalized likelihood methods, or different detector structures.

Reversal

When noise is colored, non‑Gaussian, or when the signal waveform is uncertain (unknown timing, amplitude, phase, or shape), the matched‑filter solution must be modified (pre‑whitening, generalized matched filters, GLRTs, or non‑linear detectors) and may no longer be strictly optimal.

Boundary

Applies to linear detection of a known deterministic waveform in additive white Gaussian noise when maximizing linear output SNR is the design objective. It does not directly apply to unknown or random signals, non‑additive noise models, or problems where other cost functions (e.g., false‑alarm weighted losses) dominate.

Semantic Tension

Optimality under AWGN ↔ Robustness to model mismatch: matched filtering is optimal for the AWGN model but can be fragile to deviations (coloring, interference, model uncertainty) that require alternative or adaptive detectors.

Synthesis

The matched filter provides a principled, implementable optimum for detection of known waveforms in AWGN by maximizing output SNR, but practical receivers must test and adapt to departures from the AWGN/deterministic‑signal assumptions to maintain near‑optimal performance.