Skip to Main Content
Perturbations in the sequences of pitch periods and amplitudes of the signals picked up by a throat contact microphone are investigated for the purpose of the detection of laryngeal pathology with special emphasis on the detection of cancer of the vocal folds. The subjects consist of 101 patients with various laryngeal pathologies, including 42 patients with cancer of the vocal folds, and 39 normal speakers for control purposes. The data are statistically analyzed with two acoustic measures: Koike's quotient and a novel index which is derived from reflection coefficients obtained by the maximum entropy method for spectral estimation of the sequences. It is shown that when the TNM classification of UICC is employed to describe the physiological status of the larynx of patients with cancer of the vocal folds, the T2-T4 cancers can reliably be detected with each of the two measures.