الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 551 |  |  |
| | | from | 551 | | |
Abstract
The purpose of the present study is to synthesis different nasals and vowels combinations in Modern Standard Arabic (MSA) by using Klatt synthesizer. Two production experiments and one perceptual experiment are conducted to fulfil this target.
The main goal of the production experiments is to estimate the input parameters of klatt synthesizer. The subsequent goal of the production experiments is to explore the role of the speaker gender and different phonetic variables in singling the input parameters. In order to achieve these goals different nasal and vowel combinations produced by five males and five female Egyptian speakers are acoustically analysed. The main finding of the production experiments is that the tested variables have a consistent consequence on some input parameters and context dependent consequence on other parameters. This consistency indicates that such parameters have the primacy over other parameters in signalling the phonetic contrast.
The purpose of the perceptual experiment is to evaluate the accuracy of Klatt Grid formant synthesizer in producing natural and intelligible different vowels and nasals combinations in Arabic. All synthetic syllables are created by writing Praat script. The synthesized stimuli are randomized and presented to 30 listeners in identification and discrimination tasks. With regard to perceptual results, the synthesized stimuli showed natural and intelligible nasals and vowels except for /n/ in female voices. The total naturalness percentage of synthesized syllables for males was 91.4% and 84.5% for different vowels with /m/ and /n/ respectively. Moreover, the total naturalness percentage of synthesized syllable for females was 89.06% and 52.83% for different vowels with /m/ and /n/ respectively. Regarding identification task results, the intended // was perceived correctly more than the intended //. The results showed accurate identification percentage for each vowel. In general, incorrect responses were scatter among the length counterparts of the target vowels and their neighbouring classes in acoustic space. The length counterparts of the target vowel elicited higher confusion. The vowel perception showed a symmetric pattern of confusion for the intended vowel. The intended long vowels exhibited highest percentages of correct responses than their short counterparts. The intended low vowels had the highest percentage of confusion with their length counterparts than the other vowels. The intended low vowels had highest percentage of confusion with their length counterparts than other vowels. The listener responses exhibited highly accurate gender identification than both nasal and vowel identification. There is a general trend that naturalness rates had a higher value than the identification rates of most confusable segment. This indicates that in some cases participants failed to differentiate between the synthesized and natural stimuli although they misidentified it. This tendency did not extend to the synthesized stimuli formed from /n/ and front vowels in female group.