Prediction of Postoperative Voice by Speech Synthesis in Benign Laryngeal Diseases / 대한이비인후과학회지

ABSTRACT

BACKGROUND AND OBJECTIVES: Though patients who have undergone surgery due to pathologic voice with benign laryngeal diseases are concerned about postoperative voice quality, there was no way to propose postoperative voice objectively. For this reason, the authors studied to synthesize predictive postoperative voice based on preoperative voice. MATERIALS AND METHOD: The authors evaluated 47 patients who experienced laryngeal microsurgery due to pathologic voice with benign laryngeal diseases. The voice was analysed by Computerized Speech Lab 4300B. Linear Prediction and Pitch Synchronous Overlap and Add methods were used to synthesize the predictive voice. Assessments for the synthetic voice were sound waves, spectrographic patterns with preoperative voice, and an acoustic evaluation of the postoperative voice. RESULTS: Synthetic voice showed improvement of noise component in a high frequency range that was seen in preoperative voice on spectrographic analysis. In the perceptual test, the degree of similarity in both postoperative and synthetic voice was similar and almost the same in 75% of test voice. CONCLUSION: The synthesized voice from this program was not completely identical to the real postoperative voice, but most of the tested synthetic voice was satisfactory in the perceptual test. So we conclude that this study is meaningful as a first trial that showed the possibility of synthesizing a postoperative voices by using its preoperative voice.