Fifth International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2007)

Florence, Italy
December 13-15, 2007

Fuzzy Wavelet Packet Based Feature Extraction Method Applied to Pathological Voice Signals Classification

B. S. Aghazadeh (1), H. Khadivi Heris (1), H. Ahmadi (2), M. Nikkhah-Bahrami (1)

(1) Department of Mechanical Engineering; (2) Department of Electrical Engineering
Tehran University, Tehran, Iran

In this paper an efficient fuzzy wavelet packet (WP) based feature extraction method has been used for the classification of normal voices and pathological voices of patients suffering from unilateral vocal fold paralysis (UVFP). Mother wavelet function of tenth order Daubechies (d10) has been employed to decompose signals in 5 levels. Next, WP coefficients have been used to measure energy and Shannon entropy features at different spectral sub-bands. Consequently, to find discriminant features, signals have been clustered in 2 classes using fuzzy c-means method. The amount of fuzzy membership of pathological and normal signals in their corresponding clusters is considered as a measure to quantify the discrimination ability of features. Thus, considering this measure, an optimal feature vector of length 8 has been chosen to discriminate pathological voices from normal ones. Feature vector obtained by considering nodes’ discriminant ability with classification percentage of 100 has a better performance in comparison with the feature vector including equal portion of nodes for the features of energy and entropy with the approximate classification percentage of 96. The simulation results show that fuzzy WP based feature extraction is an effective tool in voice signal analysis.
Index Terms. Voice disorders, feature extraction, wavelet packets, fuzzy sets

Full Paper (reprinted with permission from Firenze University Press)

Bibliographic reference.  Aghazadeh, B. S. / Khadivi Heris, H. / Ahmadi, H. / Nikkhah-Bahrami, M. (2007): "Fuzzy wavelet packet based feature extraction method applied to pathological voice signals classification", In MAVEBA-2007, 191-194.