Pattern Classification in Recognising Idgham Maal Ghunnah Pronunciation Using Multilayer Perceptrons
Hasliza Abu Hassan, Muhammad Nur Azan Abd Ghani, Azlee Zabidi, Wan Nazirah Wan Md Adnan, Juwairiyah Abdul Rahman, Izyani Mat Rusni · 2024
Pattern recognition is crucial in fields like speech recognition and language processing. In Arabic phonetics, identifying Idgham Maal Ghunnah pronunciation, where the nasal sound "n" assimilates into the following consonant, is a challenging task. This study introduces a method using Multilayer Perceptrons (MLPs) for Idgham Maal Ghunnah pronunciation recognition. MLPs are artificial neural networks known for their excellent pattern recognition abilities. The approach involves training an MLP model with a curated dataset of Arabic speech samples containing both instances of Idgham Maal Ghunnah pronunciation and non-Idgham instances. Successfully applying MLPs for Idgham Maal Ghunnah classification has practical implications for speech recognition systems, language learning tools, and automatic pronunciation evaluation. The research accurately distinguishes between accurate and inaccurate Idgham Maal Ghunnah pronunciation, demonstrating precision in recognizing correct articulation.