A comparative study of MFCC-KNN and LPC-KNN for hijaiyyah letters pronounciation classification system

Abstract

Reciting Al-Qur'an sometimes becomes hard to do for Indonesian because Al-Qur'an was written in Arabic which is not the native language of Indonesian. The common mistake for Indonesian is pronouncing the Hijaiyah letters. In this paper, we propose to utilize the ability of Speech Recognition to help people learn reciting Al-Qur'an in the right way. This system is built using K-Nearest Neighbor (KNN) Algorithm as the classifier. For the extraction feature, we use Linear Predictive Coding (LPC) and Mel-Frequency Cepstrum Coefficients (MFCC) and compare both. We also compare the result for system with Principal Component Analysis (PCA) and without PCA. The best result when we use LPC is 78,92% and when we use MFCC is 59,87%.