Résumé
BACKGROUND AND OBJECTIVES: In this paper, rehabilitation of voice after total laryngectomy has been suggested through the correct estimation and simulation of patients' vocal tract. Material and Methods: The author studied the shape of vocal tract during the phonation of five Korean vowels /u, o, a, e, i/ in tracheoesophageal shunt patients by magnetic resonance images (MRI). The same vocal tract was determined in each vowels from MRI. First, speech data produced by them were analyzed and also checked for speech intelligibility. Then the author tried to synthesize vowels from the vocal tract area of each vowels and from the expanded pharyngeal section of the vocal tract. RESULTS: The obtained results were as follows: 1) The sounds of /a/, /e/, /i/ were similar to natural sounds in actual patients' speech. The sound of /o/ was heard as /sigma/. The sound of /u/ was heard as strained /u/. 2) The synthesized vowels of /a/, /e/ from MRI were heard as natural sounds. The sounds of /u/, /o/, /i/ were heard as other sounds. 3) The synthesized vowel by the expanded pharyngeal section of 3 times in vowel /o/ was more naturally heard than that of 2 times. The synthesized vowel from Formfrek was more naturally heard than that from AreatoFormant. CONCLUSION: In conclusion, some of the synthesized sounds from MRI disagrees with the actual sounds produced by the subjects. This could be best identified by the synthesis from the area data. Future MRI studies should consider this problem for more accurate measurements. Also, pharyngeal areas with varied sizes should be experimented to secure better speech output because the correct shapes of vocal tract ensures correct vowel pronunciation.