Effect of Functional Endoscopic Sinus Surgery on Voice and Speech Recognition

• Published in: Journal of voice Vol. 34; no. 4; pp. 650.e1 - 650.e6
• Main Authors: Hernández-García, Estefanía, Moro-Velázquez, Laureano, González-Herranz, Ramón, Godino-Llorente, Juan Ignacio, Plaza, Guillermo
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.07.2020
• Subjects: Acoustic analysis; Acoustics; Adult; Automatic identification; Chronic Disease; Endoscopy; Female; FESS; Formants; Humans; Machine Learning; Male; Middle Aged; Nasal Polyps - physiopathology; Nasal Polyps - surgery; Pattern Recognition, Automated; Prospective Studies; Rhinitis - physiopathology; Rhinitis - surgery; Sinusitis - physiopathology; Sinusitis - surgery; Sound Spectrography; Speech Production Measurement; Speech Recognition Software; Time Factors; Treatment Outcome; Vocal Cords - physiopathology; Voice Quality; Formants; Acoustic analysis; FESS; Automatic identification
• ISSN: 0892-1997; 1873-4588
• DOI: 10.1016/j.jvoice.2019.02.012

Functional Endoscopic Sinus Surgery (FESS) is the surgery of choice for nasal polyposis and chronic rhinosinusitis. The aim of our study is to assess the influence of this surgery in the acoustic parameters of voice, and their implications in the systems of identification or verification of the speaker through the speech. A prospective study was performed between January 2017 and June 2017 including two groups of patients: those undergoing FESS, and a control group. Demographic data and GRBAS assessment were statistically analyzed. In addition, a recording of patients’ voices was made with a subsequent acoustic analysis and automatic identification of the speaker through machine learning systems, establishing the equal error rate. Samples were taken before surgery, 2 weeks after surgery and 3 months later. After FESS, a significant difference was observed in Grade, Roughness, Breathiness, Asthenia, Strain (GRBAS). Besides, acoustic analysis showed a significance decrease in fundamental frequency (F0), when compared with the control group. For the automatic identification of the speaker through computer systems, we found that the equal error rate is higher in the FESS group. Results suggest that FESS produce a decrease of F0 and changes in the vocal tract that derive in an increase in the error of recognition of the speaker in FESS patients.