Feature Extraction using Gaussian-MFCC for Speaker Recognition System

Voice is one of biometrics that is interesting to be analyzed. A voice from a person has a unique form. It cannot be identically produced twice or more. To extract this uniqueness, feature extraction is needed. One of the popular feature extractions is MFCC. The original MFCC uses triangular filter as its filter-bank. In this paper, the original filter-bank is compared with Gaussian filter-bank to obtain a better recognition output. For the decision, Euclidean distance and Manhattan distance are used. In this paper, the result shows that Gaussian filter-bank can substitute the original filter-bank of MFCC to reach a better result.