虚拟声在扬声器重放时需先通过串声消除系统(cross-talk cancellation system,CCS)来预校正;但该系统对倾听者头部位置的变化非常敏感.提出了一个基于最小最大准则的优化算法,通过在频域中自适应地调整相关参数,来达到全局的串声消除性能的最优.仿真结果表明相较常规的针对固定听音位置的CCS算法,提出的算法在头部移动时对重放声场控制具有更好的鲁棒性;尤其是在语音信号能量主要集中的低频和中频段时,具有更好的应用性.
It is an efficient way to eliminate "cross-talk" signals in spatial audio reproduction by using crosstalk cancellation system (CSS),but CCS is quite sensitive to the listener' s head movement.A global optimal algorithm is proposed to provide increased robustness for the head movement,which adjusts the parameter based on the Minimax criterion.Simulation results demonstrate the new algorithm has better robustness of reproduction than that of traditional CCS algorithm for sweet point,especially in the low and middle frequencies where the speech energy mainly gathered.