Non-linear convolution applied to acoustic signals

The acoustic evaluation of auditorium or concert halls could be retrieved by mean of auralisation. In the last 10 years, a few of 3D auralisation technique have been developed and utilised in order to evaluate the spatial sound characterisation of theatres. These techniques are: the Ambisonics method, the Stereo-Dipole method and the Wave Field Synthesis method. On the basis of all these techniques there is the convolution between a dry, not-encoded signal, and a set of properly defined (or measured) impulse responses. The convolution between the dry signal and the set of properly defined impulse responses represent the virtual spatial sound description of the room. Although the convolution method is rather old, and the spatial sound representation of the sound field goes back to early 2000’s, the convolution technique could be enhanced if the non-linear behaviour of sound sources (i.e. musical instruments) could be properly measured and emulated. In this paper, a new method of performing the convolution is presented. It is based on the Wiener- Hammerstein model and realised by means of a new numerical algorithm that utilises the Volterra series description of non-linear system. A special VST plug-in has been realised for real-time calculation and is presented in the paper.