Abstract: In this study, we investigate the implementation of correlation filter for lower face verification with different expression of images for each speaker. The motivation to implementing lower face images instead of face images is because the mouth is subject to fundamental changes during speech. Furthermore, the smaller size of a lower face image compared that of a face can reduce storage capacity and increase the speed of computation. The performance of lower face verification using Minimum Average Correlation Energy (MACE) filter is evaluated. The results are promising and offer good potential for lower face verification compare with face verification performance.