Detection of various diseases by using formant track extraction and pitch contour analysis

摘要

Pitch and formant are two of the most important characteristics of voice signals. These features can be analyzed to detect the condition of patients having Obstructive Sleep Apnea Syndrome (OSAS) and Unilateral Vocal Cord Paralysis (UVCP) before or after treatment. In this paper, we propose a method to determine whether the patients have healed from the disease or not, by using formants track extraction and pitch contour analysis. The formant frequencies are computed by using Linear Predictive Coding (LPC) method and the pitch frequencies are determined by the same LPC & calculating the real cepstrum of the voice signals. A total of 2 patients, one having OSAS and one having UVCP, were analyzed. Significant differences are found in the variation of formant frequencies of both groups. The formant frequencies of patients are higher than the formant frequencies of normal subject and the pitch frequencies are inconsistently unstable than the normal subject. The results are commended from the view of pathophysiological aspect.