In computer data storage, partial-response maximum-likelihood (PRML) is a method for converting a weak analog signal from the head of a magnetic disk or tape drive into a digital signal. PRML attempts to correctly interpret even small changes in the analog signal, whereas peak detection relies on fixed thresholds. Because PRML can correctly decode a weaker signal, it allows higher density of data recording.
For example, PRML would read the magnetic flux density pattern "70, 60, 55, 60, 70" (where 60 is the baseline signal) as binary "101", and the same for "45, 40, 30, 40, 45" (baseline of 40), whereas a peak detector would decode everything above 50 (for example) as high, and below 50 as low, so the first pattern would read "111" and the second as "000".
In the presence of colored stationary and nonstationary data-dependent noise, the performance of the PRML detector can be improved by embedding a noise prediction/whitening process into the computation algorithm of the PRML detector. This noise-prediction-based sequence-estimation framework is known as noise-predictive maximum-likelihood detection (NPML).