A multi-objective problem-solving technique for designing a two-channel quadrature mirror filter bank is proposed, where the objectives are to minimize the errors of the passband, stopband and transition band. An evolution-based algorithm, “diversity-driven multi-parent evolutionary algorithm with adaptive non-uniform mutation” (DDMPEA), is proposed for this purpose. The proposed algorithm employs the concepts of population space aggregation and fitness variance to guide the solution away from local optima. This algorithm is also validated on benchmark optimization problems. Furthermore, Wilcoxon’s test for statistical analysis at a 5% significance level confirms the effectiveness of the algorithm. From Wilcoxon’s rank-sum test, it is clear that for all considered benchmark functions, the DDMPEA is superior to other state-of-the-art optimization algorithms. The results achieved from the designed filter are compared with other available results from the existing literature. The percentage improvements in peak reconstruction error, attenuation of the stopband, and overall amplitude distortion are calculated for filter lengths of 32 and 48.