Back to list

Investigation of psychoacoustic principles for automatic mixdown algorithms

Authors Aichinger, P.
Year 2009
Thesis Type Diploma thesis
Topic Psychoacoustics
Keywords psychoacoustics, hearing model, auditory masking, masking threshold, perceptional model, timbre, discrimination of instruments, loudness, automatic mixing, identification threshold
Abstract The diploma thesis at hand describes the implementation of a psychoacoustic hearing model for modeling spectral loudness of sounds in noise, i.e. mixed instrumental sounds. The model accounts for all know physical and physiological effects between the sound source and the cerebral stage. It is used to visualize the spectral behavior of mixed instruments. It is also used to design a perceptual experiment that investigates the subjects’ ability to identify different instruments in masking environment. Different levels of masking are indicated with the aid of the new defined measure ”loudness quotient” that relates masked to unmasked (original) loudness. Instruments with a quotient of 10 % or smaller or with a residual loudness of 1 sone or less are identified to be critical in mixdowns because they cannot be identified adequate. A linear regression model is defined and proved to model the relation between the quotient of loudness and the identification probability of mixed instruments. Beside that a new approach for predicting binaural unmasking is proposed and tested on empirical data. The results of the study can be used for optimizing existing automatic mixdown algorithms, as groundwork for developing new mixing algorithms or for the revise of ordinary mixing console approaches and user interface concepts.
Supervisors Sontacchi, A., Eckel, G.