Abstract
In general, the phonemes of any language can be classified into two categories: vowels that contain no major air restriction through the vocal tract, and consonants that involve a significant restriction, and are therefore weaker in amplitude, and often "noisier," than vowels. Despite the importance of analyzing vowel phonemes in Arabic language, not a lot of research exist in the published literature yet. Consequently, this study is concerned specifically with the analysis of vowels in the Modern standard Arabic (MSA) dialect. The values of the first three formant frequencies and their derivatives in these vowels are analyzed for the purpose of automatic recognition of the six vowels of MSA through the use of an artificial neural network (ANN) based recognition system. In this paper network was tested with both one and two hidden layers, each with varying numbers of neurons. As an outcome of our experiments with four network architectures, we were able to compare, analyze and discuss the outcomes of these four network architectures.