Crosstalk cancellation using multiple synthesis sources

Abstract

The technique of crosstalk cancellation uses two synthesis loudspeakers in an attempt to produce well separated signals in a listener's ear canals. A variation on the technique can be applied in precise psychoacoustical experiments by using probe tubes in the ear canals throughout the presentation. The probe-tube application exhibits remarkable immunity to amplitude and phase variations in the responses of all transducers because the same system is used for presentation and for initial calibration. The self-correcting nature of the method even confers immunity to variations in the depth of the probe tubes within the ear canals. Because the solution to the 2-ear-2-speaker problem involves the inverse of a 2x2 matrix, the solution can become unstable, leading to pathologically large amplitudes. This problem can be almost entirely alleviated by using more than two synthesis speakers, and solving the resulting underdetermined inverse problem through the Moore-Penrose pseudoinverse. Random perturbation calculations show that using three or four synthesis speakers also reduces the ear-canal amplitude and phase sensitivity to inadvertent listener movements. However, more realistic calculations for inadvertent head-rotations indicate an additional important role for synthesis speaker location. [Work supported by the AFOSR.]