Glottal Source biometrical signature for voice pathology detection
Speech Communication
Voice pathology detection by vocal cord biomechanical parameter estimation
NOLISP'05 Proceedings of the 3rd international conference on Non-Linear Analyses and Algorithms for Speech Processing
Hi-index | 0.00 |
This paper demonstrates that linear discriminant analysis using aerodynamic and acoustic features is effective in discriminating speakers with vocal-fold nodules from normal speakers. Simultaneous aerodynamic and acoustic measurements of vocal function were taken of 14 women with bilateral vocal-fold nodules and 12 women with normal voice production. Features were extracted from the glottal airflow waveform and peaks in the acoustic spectrum for the vowel /ae/. Results show that the subglottal pressure, air flow, and open quotient are increased in the nodules group. Estimated first-formant bandwidths are increased, but result in minimal change in the first-formant amplitudes. There is no appreciable decrease in high frequency energy. Speakers with nodules may be compensating for the nodules by increasing the subglottal pressure, resulting in relatively good acoustics but increased air flows. The two best features for discrimination are open quotient and subglottal pressure.