Multirate adaptive filtering for immersive audio

This paper describes a method for implementing immersive audio rendering filters for single or multiple listeners and loudspeakers. In particular, the paper is focused on the case of single or two listeners with different loudspeaker arrays to determine the weighting vectors for the necessary FIR and IIR filters using the LMS (least-mean-squares) adaptive inverse algorithm. It describes transform-domain LMS adaptive inverse algorithm that is designed for crosstalk cancellation necessary in loudspeaker-based immersive audio rendering. Specifically, each weighting vector of the inverse filter is generated based on psychoacoustic critical band filters and uses the LMS adaptive inverse algorithm to improve performance in the sensitive frequency bands. We also investigate the sensitivity of the listening position under different number of listeners and loudspeakers with various loudspeaker geometries. Performance is measured based on the ipsilateral signal to contralateral signal (crosstalk) ratio that results from the different filter types with and without psychoacoustic critical band filtering.