Title :
Speech parameter selection for emotional stress characterization in women
Author :
Rodellar-Biarge, V. ; Palacios-Alonso, D. ; Nieto-Lluis, V. ; Gomez-Vilda, P.
Author_Institution :
Centro de Tecnol. Biomed. & Escuela Tec. Super. de Ing. Informaticos, Univ. Politec. de Madrid, Pozuelo de Alarcon, Spain
Abstract :
A total of 68 acoustical, glottal and biomechanical parameters were extracted from neutral and stressed speech from 31 women. The stress was elicited by asking the speaker to defend her false opinion in a controversial topic. The importance of the parameters was evaluated accordingly to Pearson´s coefficient, and the most relevant ones were tested in a SVM classification algorithm. We have found out that the most relevant parameters to characterize stress are pitch and the biomechanical parameters of body mass and body stiffness of the vocal folds.
Keywords :
speech processing; support vector machines; Pearson coefficient; SVM classification algorithm; biomechanical parameters; body mass; body stiffness; emotional stress characterization; glottal parameters; neutral speech; speech parameter selection; stressed speech; vocal folds; Biomechanics; Cepstral analysis; Classification algorithms; Speech; Speech processing; Stress; Support vector machines; Glottal source; biomechanical parameters; emotional stress; speaker biometry; voice production;
Conference_Titel :
Bio-inspired Intelligence (IWOBI), 2014 International Work Conference on
Conference_Location :
Liberia
DOI :
10.1109/IWOBI.2014.6913932