Intelligent algorithms for optical track audio restoration

Abstract

The Unpredictability Measure computation algorithm applied to psychoacoustic model-based broadband noise attenuation is discussed. A learning decision algorithm based on a neural network is employed for determining audio signal useful components acting as maskers of the spectral components classified as noise. An iterative algorithm for calculating the sound masking pattern is presented. The routines for precise extraction of sinusoidal components from sound spectrum were examined, such as estimation of pitch variations in the optical track audio affected by parasitic frequency modulation. The results obtained employing proposed intelligent signal processing algorithms will be presented and discussed in the paper. © Springer-Verlag Berlin Heidelberg 2005.