ISCA Archive MAVEBA 2005
ISCA Archive MAVEBA 2005

A methodology to evaluate pathological voice detection systems

Nicolás Sáenz-Lechón, Juan Ignacio Godino-Llorente, Víctor Osma-Ruiz, Pedro Gómez-Vilda, Santiago Aguilera-Navarro

This paper describes some methodological issues to be considered when designing systems for automatic detection of voice pathology, in order to allow comparisons with previous or future experiments. The proposed methodology is built around Kay Elemetrics voice disorders database, which is the only one commercially available. Discussion about key points on this database is included. Any experiment should have a cross-validation strategy, and results should supply, along with the final confusion matrix, confidence intervals for all measures. Detector performance curves such as DET plots are also considered. An example of the methodology is provided, with an experiment based on short-term parameters and Multi-layer Perceptrons.

Index Terms. Voice pathology detection, pathological voice databases, cross-validation, Multi-Layer Perceptrons.