Noise Reduction in Automatic Detection of Hypernasality in Children

Abstract

In this paper a methodology to reduce the background noise in a hypernasality detector system using spectral subtraction method is presented, some classical measures of quality and intelligibility are used to evaluate the speech enhancements algorithms used in the system. A linear classifier is used for the hypernasality detection and the results obtained with different spectral subtraction algorithms are compared. The results show that the spectral subtraction techniques can be used to improve the performance of the classifier in the detection of hypernasality when signals are contaminated with additive noise.

Keywords

• Reducción de ruido
• medidas de calidad
• sustracción espectral
• mejora señales voz.