Crosstalk cancellation system (CCS), which plays an important role in a 3D audio system applied in different fields including virtual reality, is non-robust to some disturbances caused by the listener’s head movement or rotation. To deal with the robustness of designing crosstalk cancellation filters against these slight disturbances is still a chanllege.

The research team led by Prof. LI Junfeng and Dr. XIA Risheng at the Institute of Acoustics of Chinese Academy of Sciences proposed a stochastic robust approximation crosstalk cancellation method based on random perturbation matrix, to enhance the performance of crosstalk cancellation in 3D audio reproduction via loudspeakers. The study was published in Chinese Journal of Electronics.

Mathematically, the crosstalk cancellation can be considered as an approximation problem with basic objective in the different forms, such as Minmax and Least mean squares (LMS). In practical applications, the transfer function matrix is unavoidably influenced by some perturbations and errors due to misalignments, tiny head movement and so on.

Instead of adopting common design criterion for crosstalk cancellation, researchers introduced random variable and used mathematical expectation as the objective function in this study, converting the problem of crosstalk cancellation to the stochastic robust approximation problem.

From a statistical point of view, changes in the transfer function caused by all the uncertain disturbances or errors can be modeled as random variables subject to a certain distribution. A random variable matrix was introduced to characterize the variations of the transfer function matrix between the loudspeakers and the listener resulted from listener’s tiny head movement.

Channel separation (CHS), commonly used as the evaluation measure for crosstalk cancelation performance, was computed in experimental simulations. Compared with traditional crosstalk cancellation methods, experimental simulation results demonstrated that this stochastic approximation method can improve the crosstalk robustness, especially when the nominal transfer matrix is ill-conditioned.

The research team has focused on crosstalk cancellation in 3D audio for many years and Prof. LI Junfeng's research mainly concentrates on psychoacoustics, speech signal processing and 3D audio technology.

This study was supported by the National Basic Research Program of China (973 Program), the National Natural Science Foundation of China, and the Strategy priority Research Program of the Chinese Academy of Sciences.