Sound synthesis methods of fire flame sound for predicting hidden images

Abstract

A methodology is proposed and investigated for realistically synthesizing aerodynamic sounds, such as fire sounds based on a graphically generated animation of fire. The proposed technique, “bandwidth extension method,” is a physically based combustion sound model for synthesizing low frequency flame sounds from a visual flame simulation, which runs at low temporal sampling rates. The spectral bandwidth approach uses noise matching combustion sound spectra, whereas the data-driven texture approach uses input flame sound recordings. Various simulations of fire animations are presented, and a comparison is provided of small- and large-scale phenomena for plausible synthesized and recorded flame sounds of various flames.