Speaker
Description
This work presents a formulation for predicting Crosstalk Cancellation (CTC) performance in the presence of stochastic additive perturbations to the plant matrix. The reproduction of binaural signals over loudspeakers using CTC is highly sensitive to plant modelling errors, arising from simplified models (e.g., neglecting loudspeaker directivity or head scattering) as well as unpredictable factors such as environmental changes or listener–loudspeaker misalignment. These effects introduce a mismatch between the assumed and actual plant matrix, degrading the accuracy of the reproduced binaural signals and the perceived spatial audio. Regularisation provides a means to stabilise the inverse solution, but it may reduce the achievable CTC performance in ideal conditions. This work derives a frequency-domain formulation, based on Singular Value Decomposition, to predict both the expected loss in CTC performance and the maximum achievable CTC level under stochastic additive perturbations when filters are designed using Tikhonov regularisation. Numerical simulations demonstrate the accuracy of the proposed approach and its validity under realistic perturbation conditions.