Using Artificial Neural Networks to generate virtual acoustic reality applied on escape training in blind conditions

Abstract:

This work presents a new approach to obtain the Binaural Impulse Responses(BIRs) to be applied in auralization systems, by using Artificial Neural Networks(ANN). The main goal is to implement the spectral modification of a Head-Related Impulse Response(HRIR) by using a neural network, instead of performing traditional signal processing such as convolution through Fourier transforms. The input data for the trained networks are acoustic rays carrying the power spectrum and the arriving direction. These rays were obtained with the hybrid method implemented at the acoustic simulation program RAIOS. The tests performed over a subset of HRIR directions show that the new method is capable to substitute with advantages the traditional procedure and to generate BIR's components with very small errors and much lower computational effort than the traditional method. Time and frequency domain results are …

Año de publicación:

2010

Keywords:

Fuente:

google