Comparison of cepstral coefficients to other voice evaluation parameters in patients with occupational dysphonia
More details
Hide details
Instytut Medycyny Pracy w Łodzi / Nofer Institute of Occupational Medicine, Łódź, Poland (Klinika Audiologii i Foniatrii / Audiology and Phoniatrics Clinic)
Politechnika Łódzka / Lodz University of Technology, Łódź, Poland (Instytut Elektroniki / Institute of Electronics)
Uniwersytet Łódzki / University of Lodz, Łódź, Poland (Wydział Ekonomiczno-Socjologiczny, Katedra Statystyki Ekonomicznej i Społecznej / Faculty of Economics and Sociology, Chair of Economic and Social Statics)
Ewa Niebudek-Bogusz   

Klinika Audiologii i Foniatrii, Instytut Medycyny Pracy w Łodzi, ul. św. Teresy 8, 91-348 Łódź
Med Pr 2013;64(6):805–816
Background: Special consideration has recently been given to cepstral analysis with mel-frequency cepstral coefficients (MFCCs). The aim of this study was to assess the applicability of MFCCs in acoustic analysis for diagnosing occupational dysphonia in comparison to subjective and objective parameters of voice evaluation. Materials and Methods: The study comprised 2 groups, one of 55 female teachers (mean age: 45 years) with occupational dysphonia confirmed by videostroboscopy and 40 female controls with normal voice (mean age: 43 years). The acoustic samples involving sustained vowels "a" and four standardized sentences were analyzed by computed analysis of MFCCs. The results were compared to acoustic parameters of jitter and shimmer groups, noise to harmonic ratio, Yanagihara index evaluating the grade of hoarseness, the aerodynamic parameter: maximum phonation time and also subjective parameters: GRBAS perceptual scale and Voice Handicap Index (VHI). Results: The compared results revealed differences between the study and control groups, significant for MFCC2, MFCC3, MFCC5, MFCC6, MFCC8, MFCC10, particularly for MFCC6 (p < 0.001) and MFCC8 (p < 0.009), which may suggest their clinical applicability. In the study group, MFCC4, MFCC8 and MFCC10 correlated significantly with the major objective parameters of voice assessment. Moreover, MFCC8 coefficient, which in the female teachers correlated with all eight objective parameters, also showed the significant relation with perceptual voice feature A (asthenity) of subjective scale GRBAS, characteristic of weak tired voice. Conclusions: The cepstral analysis with mel frequency cepstral coefficients is a promising tool for evaluating occupational voice disorders, capable of reflecting the perceptual voice features better than other methods of acoustic analysis. Med Pr 2013;64(6):805–816