A new approach based on Empirical Mode Decomposition for the analysis of voice features

MARÍA EUGENIA TORRES; GASTÓN SCHLOTTHAUER; HUGO LEONARDO RUFINER

Dublin, Irlanda

Conferencia; COST2102 International School Spring 2009; 2008

The European Science Foundation (ESF) funded (Co-operation in Science & Technology) COST Action

Empirical Mode Decomposition (EMD) is a data driven technique proposed by Huang. We explore properties of the intrinsic mode functions (IMFs) and present two applications to the analysis of voice.First we apply EMD to obtain new tools for discrimination between normal and pathological voices, applying it to speech signals corresponding to real and simulated sustained vowels. For the synthetic sustained vowels we propose a phonation model that includes perturbations implied in common laryngeal pathologies. By means or standard spectral analysis of the IMFs, we extract features from each signal. Using a basic pattern classification algorithm, the spectral features of only three intrinsic mode functions are enough to discriminate between normal and pathological voices, with a data base of 106 different voices, half normal and half of diverse pathologies. Additionally we consider two pathologies of different etiology and treatment, which, given the similarity of their voice characteristics, are frequently  misdiagnosed in clinical practice: muscular tension dysphonia and adductor spasmodic dysphonia.Preliminary results with a reduced real data base (33 pathological voices) suggest that this approach could provide useful orientation to physicians and voice pathologists. A second application to the extraction of the voice fundamental frequency will be introduced. A new algorithm and preliminary results will be presented. Taking into account the relationship between this parameter and emotions