EVIDENCE OF VOCAL CORD PATHOLOGY FROM THE MUCOSAL WAVE CEPSTRAL CONTENTS P. Gómez, J. I. Godino, F. Rodríguez, F. Díaz, V. Nieto, A. Álvarez, V. Rodellar DATSI, Facultad de Informática, Universidad Politécnica de Madrid, Campus de Montegancedo, s/n, 28660 Boadilla del Monte, Madrid, Spain, e-mail:pedro@pino.datsi.fi.upm.es ABSTRACT The impact of voice pathologies caused by physiological alterations of the vocal cords is becoming a very important issue due to vocal abuse and unhealthy habits. Early detection of incipient damages to the cords may help in improving the prognosis, treatment and care of these pathologies. Information derived from the speech signal may help in detecting early stages of pathology, and to prevent them by assisting experts in voice therapy to correct vocal abuse in children and in professionals depending on voice as speakers, singers or lecturers among others. The present paper is devoted to detect the presence of certain pathologies in the voice or speech signal from the cepstral contents of the mucosal wave reconstructed by inverse filtering, using conclusions derived from the behavior of a 2-m vocal cord model. This can be of application in speech and sing education, and in pathology screening. 1. INTRODUCTION The modelling of the vocal folds has been an old objective since the early works of Flanagan and Ishizaka [10] with a double purpose: to detect the vocal tract model and the glottal pulse pattern on one side, and to synthesize speech with a higher degree of naturalness on the other. Another important objective was to improve our understanding of the vocal folds operation during phonation both during normal and disordered voice production. A series of models devoted to explain the behavior of the vocal folds in 1-D, 2-D and 3-D have been produced during the last three decades which have completed our view of the phenomenon of voice production [13]. Under the purpose of the present work a model including the most relevant features have been used and adapted [8]: 2-mass assymetrical modelling, non-linear coupling between mass movement and glottal aperture, cord collision effects and non-linearities taken into account, defficient closure effects, lung flux excitation and vocal tract coupling effects. This is a modification of a well-known model through literature on the topic [3], [6]. The main characteristics of the model may be seen in Figure 1. The dynamic equations of this system are a set of four integro- differential equations, one for each of the masses in the system, similar to the one relative to mass M r1 : 0 ) ( 2 1 12 1 1 1 1 1 1 1 t t r r r r r r r r r x dt v v K dt v K dt dv M R v f (1) Ml2 Ml1 Kl1 Kl2 Kl12 Mr2 Mr1 Kr1 Kr2 Kr12 p i p o Supraglottal cavity Subglottal cavity x y Figure 1. Schematic structure of the two-mass model. a) Right cord span Left cord span t x v x t b) Right cord speed 1 1 2 2 3 34 4 Figure 2. First vibration mode of the vocal cords. a) Glottal aperture. b) Right cord speed (unidimensional). It is of most importance for our study to consider the behavior of the model with one mass per cord (with M l2 =M r2 =0 for example), as depicted in Figure 2. The vibration cycle starts at instant 1, when both cords innitiate a fast separation (in the case of the right cord toward larger values of x). At instant 2 the right and left cords (considered symmetric) have arrived to their maximum span where the speed of the cords becomes zero. From this point on, the elastic forces restore the cords to their resting position, where at instant 3 both cords collide and bounce to 4. The time that both cords remain in contact is very short, depending on the second order vibration of the model (known as mucosal wave). The intensity of the collision (the slope from 3 to 4) is of special importance to infer if phonation is overstressed, pointing to damages in the vocal cords. 2. VOCAL CORD BEHAVIOR The model can reproduce most of the features of normal or disordered voice, for instance Figure 3.a shows the glottal V - 437 0-7803-8484-9/04/$20.00 ©2004 IEEE ICASSP 2004 ➠ ➡