SIGNAL COMPACTION BY MAXIMUM VERISIMILITUDE
Dan Ştefănoiu, Janetta Culiţă · 2008
De regula, datele esantionate furnizate de un proces sunt afectate de perturbaţii. Desi inainte de prelucrarea datelor este necesara deparazitarea lor, zgomotele perturbatoare pot ingloba o parte a informaţiei dorite. In acest caz, este adecvata o deparazitare parţiala, ca etapa preliminara a procedurii de prelucrare. O alta cerinţa frecvent intâlnita in pre-procesarea datelor o constituie compresia informaţiei utile intr-un numar mai mic de date, care impreuna cu deparazitarea parţiala descriu conceptul de “compactare de semnal”. Articolul prezinta o metoda originala de obţinere a datelor preliminare compactate, bazata pe conceptul de “verosimilitate”, nefiind necesar un semnal de sincronizare care insoţeste datele achiziţionate. Sampled data provided by a process are usually corrupted by various noises. Although data are required to be denoised before applying any further processing, the corrupting noises could encode a part of desired information. A partial denoising is suitable in this case, as part of pre-processing procedure. Another requirement frequently used in data pre-processing is to compress the useful information in a smaller number of data. Partial denoising and compression are gathered in the concept of “signal compaction”. The paper introduces an original method for providing compacted preliminary data, relied on the concept of “verisimilitude” and requiring no synchronization signal accompanying the acquired data.