Room acoustics modeling using a hybrid method with fast auralization with artificial neural network techniques

摘要

This work presents a new technique to produce fast and reliable auralizations with the computer code RAIOS, a room acoustics simulator, product of research development at the Instrumentation in Dynamics, Acoustics and Vibrations Lab, State University of Rio de Janeiro. It discusses briefly the hybrid model used in the room simulation and the binaural room impulse responses generation technique using artificial neural networks, with a significant reduction in computational cost of around 85%. It is shown that the binaural impulse responses generated by the classical convolution method and the artificial neural network technique are almost indistinguishable. In the sequel, the measured and simulated binaural impulse responses for one of the rooms used in the Round Robin 4 simulation code's inter-comparison is presented, showing a quite good agreement.