Convention Paper
Real-Time Auralization Employing a Not-Linear, Not-Time-Invariant Convolver
AES Convention 123 · Paper 7295 · October 2007
Abstract
The paper reports the first results of listening tests performed with a new software tool, capable of not-linear
convolution (employing the Diagonal Volterra Kernel approach) and of time-variation (employing efficient
morphing among a number of kernels).
The listening tests were done in a special listening room, employing a menu-driven playback system, capable of
presenting blindly sound samples recorded from real-world devices and samples simulated employing the new
software tool, and, for comparison, samples obtained by traditional linear, time-invariant convolution. The listener
fills up a questionnaire for each sound sample, being able to switch them back and forth for better comparing.
The results show that this new device-emulation tool provides much better results than already-existing convolution
plugins (which only emulate the linear, time-invariant behavior), requiring little computational load and causing
short latency and prompt reaction to user’s action.
