Fast synthesis of perceptually adequate room impulse responses from ultrasonic measurements.

Audio augmented reality (AAR) applications need to render virtual sounds with acoustic effects that match the real environment of the user to create an experience with strong sense of presence. This audio rendering process can be formulated as the convolution between the dry sound signal and the room impulse response (IR) that covers the audible frequency spectrum (20Hz - 20kHz). While the IR can be pre-calculated in virtual reality (VR) scenes, AR applications need to continuously estimate it. We propose a method to synthesize room IRs based on the corresponding IR in the ultrasound frequency band (20kHz - 22kHz) and two parameters we propose in this paper: slope factor and RT60 ratio. We assess the synthesized IRs using common acoustic metrics and we conducted a user study to evaluate participants' perceptual similarity between the sounds rendered with the synthesized IR and with the recorded IR in different rooms. The method requires only a small number of pre-measurements in the environment to determine the synthesis parameters and it uses only inaudible signals at runtime for fast IR synthesis, making it well suited for interactive AAR applications.