Pattern classification in recognizing Qalqalah Kubra pronuncation using multilayer perceptrons

Abstract

Quranic recitations require precise pronunciation in its recitation. Because of this, Tajweed is important as a set of rules that govern how certain verses must be pronounced. One of the many Tajweed rules is called Qalqalah. The voice signals of Qalqalah Kubro (QK) (one of the Qalqalah variations) pronunciation have distinct patterns which can be recognized with pattern classification algorithms such as Multilayer Perceptron (MLP). This study investigates the performance of the MLP in identifying correct pronunciation of QK of a reader. The pronunciation sound waves of QK were first divided into equal length segments. Next, important features were extracted using Mel Frequency Cepstrum Coefficient (MFCC) analysis. After training, the MLP performance was analyzed to discriminate between correct and incorrect pronunciations. Results show that the MLP classifier trained using the MFCC features was able to accurately distinguish between the two cases.