Real Time Binaural Synthesis of Moving Sound Sources over Loudspeakers

Abstract

Immersive audio systems have been achieving increasing interest in the field of sound synthesis. These systems can reproduce enhanced perceived acoustic images, obtaining a better sound quality than the classical stereo reproduction. In this paper, a real time implementation of binaural reproduction over loudspeakers is proposed considering moving sound sources. The proposed system is based on an efficient 3D head related impulse responses (HRIRs) interpolation that allows to generate a moving sound source and a recursive algorithm for the crosstalk elimination to obtain a performing reproduction over loudspeakers. To achieve the HRIRs interpolation, the impulse responses are decomposed in time into early reflections and reverberant tail and the first part is then divided into two frequency bands. A peak detection and matching algorithm is reserved for the high-frequency part of the HRIRs and combined with an alignment algorithm before a linear interpolation. The other portions of the HRIRs are simply linearly interpolated and the resulting interpolated HRIR is obtained by the concatenation of early reflections and reverberant tail. Therefore, the input signal is filtered with the interpolated HRIRs and then a crosstalk elimination algorithm, called Recursive Ambiophonic Crosstalk Elimination (RACE), is applied. RACE allows to eliminate the crosstalk starting from the listener-loudspeaker distance, the distance between loudspeakers and the listener head radius and is based on an appropriate inversion, attenuation and delay of the undesired signal. The proposed system has been evaluated through several experimental tests that have confirmed its effectiveness in comparison with the state of the art.