Mel frequency cepstral coefficients (mfcc) feature extraction enhancement in the application of speech recognition: A comparison study

Mel Frequency Cepstral Coefficients (MFCCs) are the most widely used features in the majority of the speaker and speech recognition applications. Since 1980s, remarkable efforts have been undertaken for the development of these features. Issues such as use suitable spectral estimation methods, design of effective filter banks, and the number of chosen features all play an important role in the performance and robustness of the speech recognition systems. This paper provides an overview of MFCC's enhancement techniques that are applied in speech recognition systems. The details such as accuracy, types of environments, the nature of data, and the number of features are investigated and summarized in the table combined with the corresponding key references. Benefits and drawbacks of these MFCC's enhancement techniques have been discussed. This study will hopefully contribute to raising initiatives towards the enhancement of MFCC in terms of robustness features, high accuracy, and less complexity.