Back to list

Signal decorrelation for diffuse sound fields

Authors Blochberger, M.
Year 2018
Thesis Type Audio Engineering project
Topic Spatial Audio
Abstract The goal of signal decorrelation is to minimize the correlation coefficients of existing signals or to generate an arbitrary number of decorrelated signals from one source signal. This should however not introduce any disturbing amount of perceivable reverberation or other perceivable change or destruction in the resulting signals. The target method is applied for example in adaptive systems for multi-channel parametric audio processing as these often rely on decorrelated signals to work efficiently or at all. A typical goal in parametric 3D audio processing deals with the creation of diffuse sound fields for resolution enhancements of first-order or low-order Ambisonic recordings. This work confirms that filters based on randomized group delays yield good results at relatively small computational effort. Using the relation between group delay and phase the corresponding all-pass filters can be generated. These all-pass structures are used to generate an encoded higher-order Ambisonics signal with diffuseness by applying a new strategy to distribute the signals. The new algorithm is implemented in a filter design GUI and as a VST plugin and evaluated in different variations as well as compared to known algorithms.
Supervisors Zotter, F.