Study on near-field crosstalk cancellation based on least square algorithm

Crosstalk cancellation is an important issue in loudspeaker-based virtual auditory. Most of the correlated researches focus on far-field situation. This paper addresses near-field crosstalk cancellation problem, which has more challenges and seldom been addressed. Firstly loudspeaker can no longer be modeled as point source so that the head related transfer functions (HRTF) is very difficult to measure and the crosstalk problem is more complicated. Secondly any tiny user's head movement can not be neglected. In this paper, we use the boundary element calculation method with a scanned dummy-head (BHead210) model to obtain near-field HRTF data. Then by simulating loudspeaker as a combination of multiple synchronized point sources and considering possible head-position biases, a robust crosstalk cancellation method based on least square error is proposed. We evaluate the cancellation performance on system condition number, dynamic range loss and channel separation factor. The simulation results conclude that the proposed method is more suitable in near-field crosstalk cancellation task.