Spatial Auditory Blurring and Applications to Multichannel Audio Coding
Adrien Daniel · HAL (Le Centre pour la Communication Scientifique Directe) · 2011
This work concerns the telecommunications area, and more specifically the transmission of multichannel audio signals. Four psychoacoustic experiments were carried out in order to study the spatial resolution of the auditory system-also known as localization blur-in the presence of distracting sounds. As a result, localization blur increases when these distracters are present, bringing to light what we will refer to as the phenomenon of "spatial blurring". These experiments assess the effect of several variables on spatial blurring: the frequencies of both the sound source under consideration and the distracting sources, their level, their spatial position, and the number of distracting sources. Except for the spatial position of distracting sources, all of these variables have been shown to have an effect on spatial blurring. This thesis also deals with the modeling of this phenomenon, in order to predict auditory spatial resolution as a function of the sound scene characteristics (number of present sources, their frequency and their level). Finally, two multichannel audio coding schemes are proposed that take advantage of this model in order to reduce the information to be transmitted: one is based on a parametric representation (downmix + spatial parameters) of the multichannel signal and the other is based on the Higher-Order Ambisonics (HOA) representation. These schemes are both based on the original idea of dynamically adjusting the spatial accuracy of representation of the multichannel signal in a way that shapes the resulting spatial distortions within localization blur, such that they remain unnoticeable.