Random Audio Signal Processing

In this talk we will introduce elementary concepts referring to random audio signal processing. This expression should not be confused with any reference to a sound/timber property or a characterization of a signal's contents, but as a reference to signal models and hypotheses related to signal description. Deterministic and stochastic approaches for classical problems such as prediction and interpolation will be compared, and solutions for stochastic filtering problems with the methods of Wiener and Kalman will be presented. Besides, connections will be established between the introduced concepts and traditional sound processing problems, such as ADPCM compression, morphing, dereverberation and noise reduction.