Spectral envelope first peak and periodic component in pathological voices: A spectral analysis

Abstract

This paper discusses the emergence of a first peak in the spectral envelope of subjects with vocal folds pathologies and propose this peak as a feature to precocious detection of pathological voices. The relation between the spectral envelope first peak with the first and second harmonic amplitudes, the noise and the spectral tilt, together with some identification results are presented. Aperiodic/noise based features can then be used in the detection of vocal folds pathologies in a more advanced stage of the disease, while the presence of a first peak in the spectral envelope can be used in less advanced stages. In intermediate stages of the disease both features can be observed. The proposed features obtain an accuracy of 94.2% in the MEEI database and 100% accuracy in the DBSP database using simple classification systems.