Dynamic MRI for similar articulation in Arabic

Recent  advances  in  dynamic  Magnetic Resonance Imaging (MRI) have made it possible to study the  anatomy  and  dynamic  motion  of  the  vocal  tract  during speech  production in  detail. Understanding  human  speech production greatly benefits language acquisition. The Arabic language  has  novel  sounds  that  are  not  found  in  other languages.  Twelve  Arabic  letters have  similar  articulation and are very challenging to learn for native and  non-native speakers.  This  study  utilized  dynamic  MRI  scanning  to investigate  these  twelve  Arabic  letters’  anatomy  and dynamic  motion.  It  is  the  first  to  discuss  the  details  of acquiring dynamic MRI for recording Arabic phonetics. We developed  a  dynamic  MRI  protocol  for  this  study,  which effectively captured the  motion of  targeted sounds,  despite being  produced  from  different  articulation  points.  The results  showed  that  the  investigated  letters  might  sound similar. Yet, there is a vast  difference between them, which was detected clearly by comparing the outline contouring of the  MRI images.  The  results can  be a  reliable  anatomical reference to teach the difference between the target sounds. Moreover,  it  helps  us  understand  Arabic  phonetics’ intricacies  and  contributes  to  designing  more  practical, realistic, and  engaging language  learning tools  for students of all ages.

The speaker was trained to maintain a static posture during the production of each sound. Training sessions were repeated until the investigator ensured the speaker knew what he would do during the MRI scanning. The speaker was informed about the sound intended to be sustained and guided for speech initiation before each acquisition sequence using the available inter-communicator system to ensure proper speech production. The speaker started articulation after the sound of the scanning had started and repeated the articulation about five times throughout the scan. Each letter was pronounced with a glottal stop /a/ at the beginning to record the articulation point. The average recording time for each sound was measured by recording the voice of each letter separately. In this experiment, 15 seconds was the acquisition time to record each sound with a 180-second total scan duration for the twelve target sounds. The ImageJ image processing software has been widely used in various fields of scientific research for its robust image analysis and processing capabilities. In addition, it is vital in medical research, where it is used to analyze and enhance medical images such as MRI and CT scans. Our study used the software to enhance the images and clearly define the vocal tract parts’ movement. This was done by adjusting the Brightness/Contrast (B&C) to a minimum value of -83.78 and a maximum value of 606.23, resulting in a more detailed image better suited for analysis.