Cepstral, Spectral and Time-Based Analysis of Voices of Esophageal Speakers

Abstract

Objective: To evaluate esophageal speech quality in patients after total laryngectomy using cepstral, spectral and time-based measures. Patients and Methods: The subjects were all males and included 20 total laryngectomy patients and 20 age-matched normal controls. The sustained vowel /a/ was measured using the tools of MDVP (Multi-Dimensional Voice Program) and ADSV (Analysis of Dysphonia in Speech and Voice). Results: Compared to normal controls, esophageal speakers showed significantly lower values for cepstral peak prominence (CPP) and a low- to high-frequency spectral energy (L/H) ratio, but their mean CPP fundamental frequency (CPP F0) was higher. Highly negative correlations were found between CPP and jitter (percent jitter), shimmer (percent shimmer) and noise-to-harmonics ratio (NHR). Positive correlations were also found between L/H ratio and CPP. CPP F0 showed a positive correlation with F0, shimmer and NHR. In addition, CPP and L/H ratio parameters indicated an area under the curve of 0.93 or higher compared to that for time-based measures. Conclusion: This study showed that cepstral and spectral analyses are more effective than time-based analysis in examining the voice quality of alaryngeal speakers. The present findings are expected to contribute to establishing the baseline for voice quality characteristics in total laryngectomy patients.