Speech quality based on Arabic pronunciation using MFCC and LDA: Investigating the emphatic consonants

Arabic is widely considered as one of the most difficult languages which made up of phoneme-based and has its own category. In that regard, determining its speech quality is a big challenge. The letters that particularly found difficulties are the emphatics and many speakers do find articulating these letters as a challenge. The emphatic consonants are (sad), (Dhad), (Taa'), (Dzo) and (Qaf). In this study, we investigate these five letters of Arabic, which are spoken by non-native people using speech evaluating approach. Speech samples were recorded using Malaysian students and employing the International Phonetic Alphabet of Arabic chart as a reference. The recorded sound was analyzed and tested based on different coefficients of Mel-Frequency Ceptral Coefficient (MFCC) for feature extraction whilst the classifier used is a Linear Discriminat Analysis (LDA). The results showed that, among the letters, the highest accuracy rate achieved is 92.500% for (Taa'), when training for 80% of data by using 35 Coefficients of MFCC.