Identification of Nasalization (Ghunnah) in Classical Arabic Dialect Using ANN

Ali H. Meftah (King Saud Universty & College of Computer and Information Sciences, Saudi Arabia); Yousef A Alotaibi (King Saud University, Saudi Arabia)

Almost all languages contain nasalized vowels and/or consonants and there is much research in the field of classification, detection, and recognition of phonemic nasalization with different features. In Arabic language /m/ and /n/ are the nasal phonemes, but the rule of the recitation of The Holy Quran (THQ) requires the conversion or mixing of non-nasalized phonemes that come after /n/ or /m/ to produce the Ghunnah (nasalization). Our aim in this study is to classify phonemes in terms of whether they come after /n/ or /m/ or not, in order to produce a robust system used for classifying nasalized and nonnasalized phonemes. It is important to clarify our difficult goal which is the classify nasalized phoneme without dealing with inherent nasal phonemes themselves. We applied a multilayer perceptron classifier by using the first three formants. Our system accuracy was in the range 71.5 to 85.4% according to the size of the used training and testing data subsets.

Conference: UKSim-AMSS 22nd International Conference on Computer Modelling and Simulation, UKSim2020

Published: Mar 25, 2020

DOI: 10.5013/IJSSST.a.21.02.24