Abstract: In order to address the difficulties in balancing robustness and performance of conventional loudspeaker selection methods in sound zone control, a loudspeaker selection method based on the principle of minimizing the condition number of the transfer function matrix is proposed. The method is further improved by combining this strategy with the minimum mean squared error method, which not only maintains algorithm robustness but also achieves better reproduction accuracy compared to conventional iterative methods. By adjusting the threshold parameters of the improved model, a balance between robustness, computational complexity, and system performance can be achieved. Free field, reverberant field, and actual room measurement simulations, along with simulated subjective evaluation results, demonstrate that the improved condition number model has greater universality and flexibility, meeting the needs of various application scenarios.