This paper focuses on how to improve the performance of Acoustic Echo Cancellation AEC in voice communication systems. A variable step-size, dynamic affine projection algorithm named as DSVSSAPA is devised for this purpose. The proposed version combines the merits of the variable step size APA as well as uses dynamic selection of the input vectors to reduce its computational power, which has not been found in literature elsewhere. The version proposed in this paper overcomes drawbacks incorporated with the standard AP algorithm. These drawbacks are convergence speed-misalignment trade off problem and the high computational complexity of the algorithm. The proposed algorithm reduces the computational complexity by carrying the update after satisfying a specific criterion. Performance of the proposed echo canceller is evaluated and compared with those obtained from the same echo canceller but based on conventional algorithms such as the normalized least mean square algorithm NLMS and the original affine projection algorithm APA. The performance is measured in terms of misalignment and echo return loss enhancement ERLE. Simulation results showed superior ERL performance of the DSVSSAPA compared to other algorithms. It also achieves 17.73% reduction in computational complexity compared with conventional APA which is useful in portable communication applications.