Late Reverberation
Late reverberation, the lingering echoes in an acoustic environment after the initial sound, is a significant challenge in audio processing and analysis, impacting speech recognition, sound quality, and virtual acoustic rendering. Current research focuses on improving the accuracy and efficiency of modeling and manipulating late reverberation using techniques like neural networks (including convolutional and recurrent architectures), feedback delay networks, and advanced signal processing algorithms such as weighted prediction error (WPE) with data priors. These advancements aim to enhance the realism of virtual environments, improve speech intelligibility in noisy conditions, and enable more accurate room acoustic characterization for applications ranging from hearing aids to virtual reality.
Papers
Evaluation of Virtual Acoustic Environments with Different Acoustic Level of Detail
Stefan Fichna, Steven van de Par, Stephan D. Ewert
Computationally-efficient and perceptually-motivated rendering of diffuse reflections in room acoustics simulation
Stephan D. Ewert, Nico Gößling, Oliver Buttler, Steven van de Par, Hongmei Hu