Akustická analýza větného důrazu ze čtených promluv u pacientů s poruchou CNS

Abstract

Tato bakalarska prace se zabyva objektivnimi metodami hodnoceni kvality zvukovych zaznamu. Pomoci standardnich algoritmu signaloveho zpracovani a pomoci analyzy modelem slyseni byly porovnany nahravky z optickeho archivniho zaznamu na filmovem pasu se stejnymi nahravkami, ktere byly studiove upraveny procesem nazyvanym mastering. V ramci teto prace byly aplikovany algoritmy pro vypocet dynamickeho rozsahu, sirky pasma, urovne nelinearniho zkresleni, urovne impulzniho sumu a algoritmus PEMO-Q, jehoz soucasti je digitalni model slyseni. Krome analyzy urovne nelinearniho zkresleni podaly na dostupnych vzorcich vsechny pouzite algoritmy vysledky, ktere odpovidaly predpokladum, ze nahravka, ktera prosla studiovym masteringem, bude vykazovat lepsi kvality, nez neupravovany zaznam primo z optickeho media.

The goal of this Bachelor's thesis is to find and implement different algorithms of objective sound quality assessment. Algorithms used in the analysis are both from the field of standard signal processing statistics and from the field of auditory modeling. The subjects of analysis in this thesis are various short audio recordings extracted from a optical movie soundtrack. Another task is to analyze the recordings using the algorithms and evaluate the results with respect to the audio original medium. The sound recordings for comparison are of the same length and the same movie origin, but of very different content and quality, since half of the recordings were mastered in a studio and are considered reference points in the comparisons. Appart from the nonlinear distortion analysis, all the other algorithms provided very reasonable output and are suitable to be used in objective test of archived sound recordings.